Fast Scalable File Distribution Over InfiniBand
- 格式:pdf
- 大小:108.73 KB
- 文档页数:12
Data Sheet BenefitsEasy to UseOur unique clamshell and single-use Flexware®clamshell assemblies were designed to offer ultimate ease of use from preparation to production. Simply insert the assembly into the clamshell and connect the tubings.Safe & Reliable ProcessOur 100% single-use, gamma-irradiated Flexware®assemblies provide quick product changeover with reduced contamination risk and cleaning for greater efficiency and process security.The high gradient accuracy of the system enhances separation performance making it ideal for challenging processes.FlexibleEasily scale-down to lower column sizes while maintaining high accuracy to meet broader processing requirements. Ability to cover multiple scales and applications offers maximum return on investment.User-friendlyThe system was designed with operator comfort and safety in mind and features a movable screen for operator convenience. The stainless-steel surface is easy to clean and the accessible components facilitate maintenance.Mobius® Chrom 20SystemFor Purification of Clinical and Process Scale BiologicsThe Mobius® Chrom 20 system is a flexible, automated, single-use chromatography systemthat enables consistent and reliable separationand purification of mAbs, vaccines, plasma, and therapeutic proteins at clinical and process scales. The innovative single-use flowpath reduces contamination risk and gives you the confidenceto tackle any chromatography challenge, whilethe flexible design featuring a standard hardware platform that can be combined with a selection of options, empowers you to meet your unique process requirements.The Mobius® Chrom 20 system delivers optimal operational flexibility for processes up to 2000L with flow rates up to 20L/min, a 5:1 turn down ratio at process scale, and enhanced isocratic and linear gradient accuracy.The Life Science business of Merckoperates as MilliporeSigma in theU.S. and Canada.System ComponentsThe Mobius® Chrom 20 system is made up of a base system and configurable options that can be addedto design a system that best meets your purification process and challenges.The base cart is the heart of the system and houses the clamshell and automation software. The unique clamshell design ensures simple installation of the flowpaths so your process is up and running rapidly, while the optional clamshell lift enables a single technician to assemble and maintain the system.The system also shares common elements with our other unit operations, such as an intuitive software interface, assembly material (PureFlex™ film), and maintenance components, making training and validation easier across scales and applications.Bubble trap support Manifold (inlet)Pump cartSingle-use flowmeter Base cart ClamshellPre-column instrumentation Post-column instrumentationAdjustable screen Mobius® Chrom 20 System ConfigurationsThe basic configuration includes:• Clamshell• Pump cart• Base cart• Post-column instrumentationPre-column 0,2µm filter andits support for protectingthe columnPre-column instrumentationfor conductivity and pHmeasurementKeyboard support forenhanced ease of useThe system options and accessories include:Clamshell lift for one-manassembly and maintenance23Mobius ® Chrom 20 Flexware ® AssembliesSingle-use Flexware ® assemblies deliver equivalent performance to that oftraditional stainless-steel systems, while providing maximized contamination control. Assemblies minimize dead space, decreasing the risk of contamination to virtually zero, while ensuring high gradient accuracy. The assemblies are delivered gamma-irradiated, ready-to-use, and are clearly labeled,significantly reducing the risk of installation error and the challenges typically associated with installing single-use flow paths.Benefits of our 100% Single-use Flowpaths• Fast changeover and reduced contamination risk from previous batches in multi-product and multi-scale productions • Combination of clamshell andFlexware ® clamshell assembly offers easy installation compared to classic tubing • Reduced hold-up volume contributes to the enhanced accuracy of the system • Reduced number of connections versus traditional single-use assembliesdecreases the risk of integrity failure • The clamshell assemblies are 100% integrity tested during manufacturing, ensuring reliable performanceFlexware ® clamshell assembly labelsMobius ® Chrom 20 system clamshell and Flexware ® clamshell assembly4Common Control Platform ® (CCP ®) SoftwareOur CCP ® 6.8 software on Windows ® 10 provides a single software platform across our portfolio of automated systems for a familiar look and feel at each step of your process and reduced training time for operators.The software offers recipe driven processing that eliminates manual operation, reduces processvariability, and minimizes risk of errors. Process operations are easily created and managed using the recipe editor, each aspect of your process is monitored from the HMI (Human Machine Interface), and the report client provides state-of-the-art batch reports.The software is designed for GMP facilities, is in line with GAMP ® 5 recommendations for automationsoftware, and fulfills FDA guideline 21 CFR Part 11 requirements for electronic records and signatures.Mobius ® Chrom 20 system CCP ® software interface5Optimized, Robust Chromatography PerformanceExceptional gradient accuracy performanceThe system is designed to deliver high accuracy gradients using pump-based mixing over an extended flowrate range. The test results below illustrate the better than 2% accuracy obtained from 4 to 20 L/min over a mixing range of 10-90% in step and linear gradient modes. This unique performance enables enhanced separation in a fully single-use mode to meet any chromatography challenge.-2.0%0%10%20%30%40%60%70%80%90%100%50%-1.5%-1.0%-0.5%0.0%0.5%1.0%1.5%2.0%12 L/min4 L/min20 L/minM i x i n g e r r o r i n p e r c e n t a g ePercentage of Pump 1 mixingLinear Gradient Error on mixingFigure 1: Linear Gradient (%) at 4 L/min, 12 L/min and 20 L/min -2.0%0%10%20%30%40%60%70%80%90%100%50%-1.5%-1.0%-0.5%0.0%0.5%1.0%1.5%2.0%M i x i n g e r r o r i n p e r c e n t a g ePercentage of Pump 1 mixingStep Gradient Error on mixing12 L/min4 L/min20 L/minFigure 2: Step Gradient (%) at 4 L/min, 12 L/min and 20 L/minLow pressure drop over full flow rate rangeThe system is designed to minimize pressure drop across the full flow rate range. The system pressure drop is below 1 bar at max flowrate.0.04681014161820120.30.20.10.40.50.60.70.80.91.0BBT online filter bypass * Bubble TrapBBT* online filter onlineBBT and filter bypassP r e s s u r e (b a r )Flowrate (LPM)Pressure DropFigure 3: Pressure drop (bar) at 4 L/min, 12 L/min and 20 L/min6Related products to support your chromatography processQuikScale ® Chromatography ColumnsDesigned to achieve maximized throughput, ourQuikScale ® columns are easily packed to deliver optimal resolution across a wide range of chromatographic applications, accommodating all media types. The flow distributor assures full, uniform media utilization within the column, assuring reproducible and reliable separations. QuikScale ® chromatography columns offer fast packing and unpacking, predictable and scalable performances, and flow adjuster seal technology for fail-safe operation in a choice of column tube materials for any application.IsoPak™ Chromatography Process ColumnsRecognized for their proven chromatographicperformance, IsoPak™ columns provide faster, cleaner, and more consistent packing and unpacking forreproducible results. IsoPak™ columns are available in a variety of options including IsoPak™ valves, antijet, hydraulic assist, and proven flow cell distribution to provide consistent performance and scalability across a wide range of column diameters.Industry-Leading Affinity and Ion Exchange Chromatography MediasEshmuno ® ResinThis unique family of ion-exchange and affinity resins is specifically designed for highly productive downstream purification.Fractogel ® ResinThis tentacle-modified synthetic polymer resin is designed for ion exchange, hydrophobic interaction, size exclusion and metal chelate affinity.Supply Robustness and Control. Stay Ahead.Mega-trends are contributing to a boom in thebiopharmaceutical industry. As a leading partner and supplier to the industry, we continue to strengthen our supply chain to support our commitment of on-time delivery reliability and product quality to help you stay ahead.To learn more about our program, please visit our website at /StayAhead .Mobius® Chrom 20 System ServicesThe pharmaceutical and biotechnology industries are highly regulated and, to help you navigate this challenging environment, we offer a wide range of services. These services help you save time, lower costs, and comply with regulations. For your peace of mind, all our services are performed by our global experts who have an intimate knowledge of our equipment backed by decades of experience.Qualification ServicesOur qualification services are designed to makethe integration of our system into your process as seamless as possible and ensure your equipment is properly installed and functioning per your pre-defined requirements.• Factory acceptance test (FAT)• Installation qualification/operational qualification (IQ/OQ)• Full test package: this service is an alternative to standard IQ/OQ for customers who wish to have tests from FAT performed again at their site• Performance qualification support (PQ)For additional information on tests performed, please contact us to get the detailed test matrix document MK_CA6580EN.Training ServicesAppropriate training for users is not only a cGMP requirement, but it also ensures your staff has the expertise to operate and manage the system as part of your manufacturing process. Our training offering has been designed to make your staff more autonomousin managing your system and your process while saving time and money. Our training services cover system use and programming with interactive hands-on sessions and, depending on the training you select, may also include:• Installing the Flexware® assemblies• Interacting with the Human Machine Interface• Manual and automatic system operation• Troubleshooting issues• Creating and managing your own recipes• Process recommendationsThese trainings can be delivered either at your site or in our M Lab™ Collaboration Centers. Please contact your local representative or email ********************* to discuss our training offering.System Service Reliance PlansTo help you ensure optimum equipment uptime while mitigating risks, we have developed a wide rangeof services and support that allow you to select a coverage level that best fits your needs. Our System Service Reliance Plans, a complete range of services for your systems, offer priority access while ensuring your equipment is properly maintained.For additional details, please refer to the System Service Reliance Plans Data Sheet (MK_DS7881EN).Specialized ServicesCCP® Software Recipe DesignEvery process is unique and, to ensure that your system is optimized to deliver the best performance, our biomanufacturing engineers will configure your process into your own CCP® software recipe. This allows your system to run fully automatic, resulting in consistency and reduced operator error.For catalogue numbers, please refer to the ordering information at the end of the document.Repair Services and Spare PartsRepair ServicesIn the event your system experiences a problem, our worldwide engineering organization will provide on-site or repair center technical support to get you back up and running as quickly as possible.Spare PartsPurchasing spare parts directly from us is the onlyway we can guarantee that you get the right parts every time, with the same level of performance as the original. For details and ordering information, check the illustrated spare parts list (MK_BR5624EN).78Mobius ® Chrom 20 System and Flexware ® Assemblies SpecificationsGeneral System SpecificationsFlow Range In gradient mode within accuracy specifications: 240–1200 L/h In Isocratic mode: 24–1200 L/h PumpsModelQuattroflow ® 1200S x 2Maximum Pressure4 barValvesOn/Off ValvesInside Clamshell: Automatic pneumatic pistonInlet Manifold: Automatic pinch valve for 5/8" ID tubingValve control (after Bubble Trap)Pneumatic piston with pressure regulatorMechanical Specifications: System and AssembliesSystem Dimensions Pump & Base carts connected H x W x D in mm 1727 x 1715 x 800 ± 20 mm (68 x 67.5 x 31.5 in.)Net WeightBase cart with Clamshell365 kg (805 lbs)Carts frame & panels Stainless steel 304L Clamshell (rear)POM Clamshell (front)PMMA Valve PadsSiliconeCasters8 (4 free, 4 locking) Inner : Polyamide 6Rolling area : PolyurethaneConnectionsInlets (10)¾" TC Outlets, 3 Fractions + 1 waste ¾" TC Filters ¾" TC Column ¾" TC Pneumatic¾" TCOperating SpecificationsTotal Operating Time Flexware® Assemblies Validated for 24 hours operation50 cycles (50 valve openings and 50 valve closings) per valveThe integrity of the installation may be compromised if used formore than 50 cyclesValve Pads Do not exceed 1200 cycles or 6 months of operationProduct Temperature Range 2 to 30 °C (36 to 86 °F)Maximum Pressure Pump 1 Manifold0 to 2 bar (0 to 29 psi)Pump 2 Manifold0 to 2 bar (0 to 29 psi)Pump Assembly0 to 4 bar (0 to 58 psi)Bubble Trap Assembly0 to 4 bar (0 to 58 psi)Clamshell assembly0 to 4 bar (0 to 58 psi)Precolumn Filter Assembly0 to 4 bar (0 to 58 psi)Pre-column Instrumentation Assembly0 to 4 bar (0 to 58 psi)Post-column Instrumentation Assembly0 to 2 bar (0 to 29 psi)Column Assembly0 to 4 bar (0 to 58 psi)System Operating Temperature20 to 30 °C (68 to 86 °F)Operating Humidity10 to 90% (non condensing)Power and Air Supply SpecificationsPower Supply220-240VAC, 50/60 Hz, 1 phase, 6.5 A max or 100–120VAC, 50/60 Hz, 1 phase, 13.6 A maxMaximum consumption 1.6 KWPneumatic Air Supply 6 bar (87 psi) min, 10 bar (145 psi) maximum, oil freeMaximum consumption 4 L/minInstrument Specifications and PerformancesPressure Indicators PIT001PIT002PT0030–4 bar (0-58 psi) ± 0.2 bar (2.9 psi)Non-intrusiveHigh Pressure Switches PSH001PSH0024.2 bar (60 psi)Non-intrusiveSingle-Use Magnetic Flowmeters FT001FT002± 2% MV between4 L/min and 20 L/minIn-lineTemperature Sensors TE2012–30 °C (35.6 °F to 86 °F) ± 1 °C (33.8 °F)Combined with conductivity sensorSingle-use Conductivity Sensors AE102AE201AAE201B Process: 0–100 mS/cm ±3% MV + 0.4 mS/cm Cleaning: 0–50 µS/cm ±3% MV + 0.4 µS/cmIn-linepH Sensors AE103AE2023–9 pH ±0.1 pH In-lineAir Sensors XS001XS002N/A Non-intrusiveUV SensorsDo not use UV Sensors in a condensing atmosphere. Condensation may lead to erroneous sensor readings.AE203/204AF45 : 0-3 AU ±2% FSOPL: 1 or 2.5 mmWavelength: 280 nmAF46 : 0-2 AU ±2% FSOPL: 1 or 2.5 mmWavelength: Dual 254/280 nm or280/300 nmIn-line*Refers to the labels on the system hardware9Ordering Information Hardware10Hardware11For additional information, please visit /chrom-systems© 2023 Merck KGaA, Darmstadt, Germany and/or its affiliates. All Rights Reserved. Merck, the vibrant M, Millipore, Mobius, Flexware, PureFlex, Common Control Platform, CCP, QuikScale, Isopak, Eshmuno, Fractogel, and M Lab are trademarks of Merck KGaA, Darmstadt, Germany or its affiliates. All other trademarks are the property of their respective owners. Detailed information on trademarks is available via publicly accessible resources.Lit. No. MK_DS5456EN Ver. 4.03052201/2023 Merck KGaAFrankfurter Strasse 25064293 Darmstadt, GermanyFlexware®kitServices*PMFXWKITXM20 kit should be ordered separatly prior to Essential Reliance Plan。
Visit for more complete productinformation, including datasheets and specifications.32OverviewThank you for your interest in OmniStream™ R-Type , Atlona’s AV over IP platform designed for residential integration. By selectingOmniStream R-Type, and AV over IP systems in general, you’re entering into a new world of integration flexibility for designing and buildingwhole-house AV systems for your customers. You’ll benefit from the performance and reliability of conventional AV distribution (HDMI ® and HDBaseT™), but also the unprecedented convenience of using standard data network cabling and equipment. This will give you newfound freedom to fully customize systems for your clients, with more application possibilities and greater efficiency than ever before.• FLEXIBLE INPUTS AND OUTPUTS – With AV over IP , system capacity is limited only by the size of the network switch and how muchstructured CATx wiring is available in the residence. You’ll never have to worry about choosing the “right” I/O size for a matrix switcher, or an expensive system upgrade to replace the existing switcher.• UNRESTRICTED AV SIGNAL ROUTING – AV over IP (also known as networked AV ) gives you freedom to independently route video, audio, and control over the network. This means that you can easilycustomize distinct application scenarios for all three, making it much easier for you to do exactly what your customers want with their whole-house AV, home theater, and automation systems.• CENTRALIZED SYSTEM MANAGEMENT – Having all devices on the network makes it very easy to configure, monitor, and manage them from a single web browser portal, saving considerable time and expense for yourself and your clients. You can also opt to provide remote monitoring services to your customers for continuous peace of mind and proactive maintenance.The advantages of AV over IP systemsIntegrating AV distribution over standard data networks brings several important benefits:33 56 6 8 14 18 202224 26 283437 37 37 38 38 424244 4446 4648 48 4850 50 52 54 56574US: +1.408.962.0515+1.877.536.3976 (US T oll Free) International: +41.43.508.4321How AV over IP systems work – the fundamentals VideoWhole-house videoFor whole-house video, decoders are typically situated at each display in a room or moreHow AV over IP systems work – the fundamentals VideoHow AV over IP systems work – the fundamentals VideoVideo processingHow AV over IP systems work – the fundamentals VideoHDMI 2.0 and HDCP 2.2Two vital design considerations for any residential AV system are 4K capability and HDCP 2.2 compliance. Both are essential to ensure full compatibility with the newest AV sources availableVideoHDRHigh Dynamic Range (HDR)It’s currently available with streaming media services, some HDR enhanced games, and select UltraUltra-fast switching modeA special switching and video processing mode is available that greatly speeds up switching time for sources by actively managing EDID and HDCP negotiation between sources and displays. The switching is seamless with transitions approximately 25 ms at 60 fps, and a clean cut from one source to another with no visual artifacts. This is a groundbreaking feature for AV over IP systems, since fast switching has generally been challenging to achieve in IP packet-based AV systems.In ultra-fast switching mode, video output is capped at 1080p. Fast switching will apply to HD sources at 720p and 1080p, as well as UHD (3840x2160).VideoHow AV over IP systems work – the fundamentals AudioThere are many different ways to integrate audio for OmniStream R-Type systems in a residence, for audio destinations including the following:How AV over IP systems work – the fundamentals AudioHDMI embedded audioOmniStream R-Type can also de-embed audio from the sources, and route the audio to destina-tions separate from the video. As an example, this may be desirable in situations in which there is a music server installed, and the client wishes to play background music through the displays or soundbars but still have video on the display, such as for a live sporting event.How AV over IP systems work – the fundamentals AudioDistribution over whole-house audio and home theater systemsThere are two primary ways in which audio sources and OmniStream R-Type can be integrated into a whole-house audio distribution system and a home theater system:1. DECODE DOLBY AND DTS BITSTREAM AUDIO FORMATS IN THE SOURCE DEVICES ANDDISTRIBUTE 2.0, 5.1, OR 7.1 CHANNEL PCM OVER THE NETWORK.encoders can stream multi-channel PCM from a source such as a Blu-ray player orAudioAudioDistribution over whole-house audio and home theater systems2. DISTRIBUTE DOLBY OR DTS BITSTREAM AUDIO OVER THE NETWORK, INSTEAD OFDECODING IN THE SOURCES.send Dolby Atmos or DTS:X audio across the network to a home theater system. These object-oriented surround sound formats cannot be decoded within the source device,How AV over IP systems work – the fundamentals AudioPC PCSystem managementVirtual MatrixAMS features a Virtual Matrix GUI that visually maps encoders and decoders in a grid format. This is very similar to a GUI for setting matrix switcher cross-connections. By clicking and selecting icons, you can select an OmniStream 512 encoder and one or more OmniStream 521 decoders to stream video, audio, and control, together or separately to specific destinations. Behind the scenes, AMS is dynamically assigning the decoders you have selected into a multicast group with the encoder.Management, firmware updates, and monitoringIf necessary, you can also use AMS to customize a configuration for an encoder and decoder, and then duplicate the same configuration to multiple devices in the system. AMS is also an essential tool to check for firmware updates, and schedule a convenient date and time to apply them to a customer’s system. Perhaps most importantly, AMS allows you to check on the health of the system, providing notifications of any problems arising with an encoder or decoder.How AV over IP systems work – the fundamentalsControlThird-party control systemsDrivers are available for popular home control and automation systems from Crestronwell as thethroughout a residence – similar to working with a matrix switcher.OmniStream R-Type systems can be controlled via IP or RS-232. In addition to assigning AV streaming from an OmniStream 512 encoder to select OmniStream 521 decoders, a wide range of system control options are available throughAV over IP system design scenariosWhole-house video and audioOverviewA new, custom home design will include an abundance of technology, from comprehensive home automation to whole-house video and audio. Working closely with the homeowner, the residential AV integrator has identified 16 display locations and 5 zones for audio. A few of the display and audio locations are situated outdoors.AV over IP system design scenariosWhole-house video and audio plus home theaterOverviewAn AV integrator is working with a homeowner to upgrade legacy whole-house audio and video systems to a more efficient, integrated solution with all signalsrunning over IPdistribution with high quality HDR video content, and the complement of Dolby Atmos and DTS:X surround sound audio.PCAtlona OmniStream TMFully compatible with DolbyVision, HDR10, and HLG @ 60 Hz4K HDRWorks over new or legacynetwork infrastructure1 GB ETHERNETReady to go with popular network switches from Cisco,Pakedge, and othersPLUG N’ PLAYSupports IEEE 802.1x, HDCP 2.2,AES-128, and other security protocolsSECURESimple, effortless setup and configuration for video wallsVIDEO WALLSPower all encoders and decoders from a network switchwith PoEPOEVirtual auto-switching withgrouped or daisy-chainedencodersAV SwitchingDeliver ultra-fast, <25 ms AV switching (1080p output)Fast SwitchingThe only platform that streamsDolby Vision @ 60 Hz and 12-bit colorR-TypeUncompromised AV over IP performance and reliabilityThe next-generation AV distribution platform for integrators to design and build flexible, scalable, and cost-effective AV systems – with best-in-class AV over IP performance and reliability over Gigabit Ethernet.OMNISTREAM PRODUCT FAMILIES – Tailor an OmniStream networked AV system design to your specific needs:• OmniStream Pro – Networked AV designed for conference rooms, boardrooms, and divisible rooms, plus AV distribution over large networks • OmniStream R-Type – Networked AV streamlined for residential and light commercial AV distribution applications • OmniStream Audio – Audio over IP integration via Dante and AES67•OmniStream USB – Connect USB devices to PCs and peripherals across a networkSingle-Channel Networked AV EncoderA T-OMNI-111Single HDMI input4K HDR plus VESA resolutions Network security including IEEE 802.1x Local/PoE-poweredPro Dual-Channel Networked AV EncoderA T-OMNI-112Dual HDMI inputs with auto-switching 4K HDR plus VESA resolutions Network security including IEEE 802.1x Local/PoE-poweredProSingle-Channel Networked AV DecoderA T-OMNI-121Single HDMI output 4K HDR plus VESA resolutions Network security including IEEE 802.1x Audio embedding/de-embedding Local/PoE-powered Optional Dolby Vision licenseProDual-Channel Networked AV DecoderA T-OMNI-122Dual HDMI outputs 4K HDR plus VESA resolutions Network security including IEEE 802.1x Audio embedding/de-embedding Local/PoE-powered Optional Dolby Vision licensePro Dual-Channel Networked AV EncoderA T-OMNI-512Dual HDMI inputs 4K HDRLocal/PoE-poweredSingle-Channel Networked AV DecoderA T-OMNI-521Single HDMI output4K HDR PoE-poweredOptional Dolby Vision licenseTwo-Input, Two-Output IP / Analog Audio BridgeA T-OMNI-232Dual line/mic inputs Dual line outputs Balanced audioSelectable 24 V phantom power and gain settings16 Output IP to Analog Audio BridgeA T-OMNI-238Eight stereo line outputs Balanced audioSelectable output gain settingsR-Type Audio Audio A T-OMNI-311Host side USB to IP adapterA T-OMNI-324Device side IP to USB adapterExtend USB peripherals across a local area networkAtlona Management System and Atlona Velocity Control Systemuser-friendly operation.。
AA Auto AnswerAAA Authentication, Authorization and AccountingAAB All-to-All BroadcastAAL Asynchronous Transfer Mode Adaption LayerAAM Automatic Acoustic ManagementAAP Applications Access Point (DEC)AARP AppleTalk Address Resolution ProtocolAAS All-to-All ScatterAASP ASCII Asynchronous Support PackageAAT Average Access TimeAATP Authorized Academic Training Program (Microsoft)ııı.ABA Address Book Archive (dosya ad uzant s ) (Palm)ABEND Abnormal End ABI Application Binary Interface ABIOS Advanced BIOS ABIST Automatic Built-In Self-Test (IBM)ABLE Adaptive Battery Life ExtenderABM Asynchronous Balanced Mode ABR Available Bit Rate ABRD Automatic Baud Rate Detection ABT Abort ABTS ASCII Block Terminal Services AC Autocheck + Automatic Computer + Alternating Current ACAP Application Configuration Access Protocol ACC Accumulator ACCMAIL Accessing the Internet Via EmailACD Automatic Call Distribution ACDI Asynchronous Communications Device Interface ACE Access Control Encryption/Entry +ACE Adobe Certified Expert +ACE Advanced Computing Environment (SCO) +ACE Automatic Computing EngineACF Access Control Field +ACF Advanced Communications FunctionACH Automated Clearing HouseACIAS Automated Calibration Interval Analysis SystemACIS American Committee for Interoperable SystemsACK AcknowledgmentACL Access Control List + Asynchronous Connection-Less (link)ACM Association for Computing MachineryACM Audio Compression Manager (Microsoft)ACMS Application Control Management SystemACP Ancillary Control Program + Auxiliary Control ProcessACPI Advanced Configuration and Power InterfaceACR Allowed Cell RateACROSS Automated Cargo Release and Operations Service SystemACS Access + Access Control Set + Access Control System +ACS Advanced Computer System (IBM) +ACS Anti Curl System + Asynchronous Communication ServerACSS Audio Cascading Style SheetsACTS Automated Computer Time ServiceACU Automatic Calling UnitACVC Ada Compiler Validation CapacityA/D Analog to DigitalADA Automatic Data Acquisitions +ADB Apple Desktop BusADC Adaptive Data Compression (protocol) (Hayes) +ADC Add with Carry + Analog to Digital ConverterADCCP Advanced Data Communication Control Procedures/ProtocolADD Automatic Document Detection (WordPerfect)ADF Automatic Document Feeder +ADF Automatically Defined Function ııı.ADF Adapter Description File (dosya ad uzant s )ADI AutoCad/AutoDesk Device Interface (driver)ADL Address Data LatchADLAT Adaptive Lattice FilterADLC Adaptive Lossless Data Compression (IBM) +Asynchronous Data Link ControlADMACS Apple Document Management And Conrol SystemADMD Administrative Management Domain (X.400)ADO ActiveX Data Objects ADP AOLserver Dynamic Pages + Automatic Data ProcessingADPCM Adaptive Differential Pulse Code Modulation ADR Address ADS Application Development Solutions (AT&T) +ADS Application Development System +ADS Automatic Distribution SystemADSC Address Status Changed +ADSC Adobe Document Structuring ConventionsADSI Active Directory Service Interface (Microsoft) +ADSI Analog Display Services InterfaceADSP AppleTalk Datastream ProtocolADSR Attack, Decay, Sustain, ReleaseADT Abstract Data Type + Application Data TypesADT Automatic Dialing UnitAE Above or EqualAEB Analog Expansion Bus (Dialogic)AEC Architecture, Engineering, Construction.AERO Aviation Industry (Domain Name) (Internet)AES Advanced Encryption Standard AESA ATM End System Address AF Auxiliary carry Flag AFA Accelerated File Access AFC Antiferro-Magnetically Coupled (media) (IBM) +AFC Automatic Font Change + Automatic Frequency Control AFD Automatic File Distribution AFDW Active Framework for Data Warehousing (Microsoft)AFI Authority and Format Identifier AFII Association for Font Information Interchange AFIRM Automated Fingerprint Image Reporting and Match AFIS Automated Fingerprint Identification System ııı.AFM Adobe Font Metrics (dosya ad uzant s ) (Adobe Systems)AFP Advanced Function Presentation (IBM) +AFP Advanced Function Printing + AFP Appletalk Filing Protocol (Macintosh)AFS Andrew File System (protocols)AFTP Advanced/Anonymous File Transfer ProtocolAGA Advanced Graphics AdapterAGC Automatic Gain ControlAGP Accelerated/Advanced Graphics PortAGRAS Antiglare-Antireflective-AntistaticAGU Address-Generation UnitAH Authentication Header (protocol)AHS Automated Highway SystemsAI Analog Input + Artificial IntelligenceAIA Applications Integration Architecture (DEC)AIC AIXwindows Interface Composer (IBM)AIFF Audio Interchange File FormatAIIM Association for Information and Image ManagementAIM AOL Instant MessengerAIN Advanced Intelligent Network (Bell Atlantic)AIP Application Infrastruture ProviderAIR Architecture Implementation ReviewAIS Automated Information SystemsAISB Association of Imaging Service BureausAISP Association of Information Systems ProfessionalsAIX Advanced Interactive Executive (IBM)ALAT Advanced Load Address TableALC Arithmetic and Logic Circuits +Automatic Level ControlALE Address Latch Enable +Application Linking and Embedding ııALGOL Algorithmic Oriented Language (bak n z IAL)ALICE Artificial Linguistic Internet Computer EntityALINK Active Link (HTML)ALIWEB Archie Like Indexing in the Web (Internet)ALN Asynchronous Learning NetworkALP Automatic Language ProcessingALR Advanced Logic Research, Inc.ALT Alternate (mode)ALT. Alternate lifestyle (USENET Newsgroup Category)ALU Arithmetic Logic UnitAMA Automatic Message AccountingAMANDDA Automated Messaging and Directory AssistanceAMASS Archival Management and Storage SystemAMD Active Matrix Display + Advanced Micro Devices, Inc.AMI Alternate Mark Inversion + American Megatrends, Inc. +Application Messaging Interface AMMA Advanced Memory Management Architecture AMPS Advanced Mobile Phone Service AMR Audio Modem RiserANCOVA Analysis of Covariance ANDF Architecture-Neutral Distribution Format ANI Automatic Number Identification ııı.ANI Animated Cursor (dosya ad uzant s ) (Microsoft)ANN Artificial Neural Networkııı.ANN Annotations (dosya ad uzant s ) (IBM)ANOVA Analysis Of VarianceANSI American National Standards InstituteAO Analog OutputAOCE Apple Open Collaboration EnvironmentAO/DI Always On/Dynamic ISDNAOE Application Operating Environment (AT&T)AOL America OnlineAOS Add Or SubtractAP Adjunct Processor + Application ProcessorA/P Accounts PayableAPA Adaptive Packet Assembly +APA All Points Addressable +APA Arithmetic Processing Accelerator APAR Authorized Program Analysis Report (IBM)APAREN Address Parity Enable (IBM)APCUG Association of PC User GroupsAPDU Application Protocol Data Units API Application Program Interface APIC Advanced Programmable Interrupt Controller (bus) (Intel)APIS Advanced Passenger Information System APL A Programming Language (Mathematics)APM Advanced Power Management (IBM)APNIC Asia-Pacific Network Information Center (Internet)APOP Authenticated Post Office Protocol ııı.APP Application (dosya ad uzant s )APPC Advanced Program-to-Program Communications (IBM)APPI Advanced Peer-to-Peer Internetworking APPLETS Applications APPN Advanced Peer-to-Peer Networking (IBM)APPS ApplicationsAPRP Adaptive Pattern Recognition ProcessingAPS Advanced Photo System + Advanced Printing Service (IBM)+ Asynchronous Protocol Specification APSE ADA Programming Support EnvironmentAPT Address Pass Through +Advanced Parallel Technology +Automatically Programmed ToolsA/R Accounts ReceivableARA AppleTalk Remote AccessARAG Antireflective-AntiglareARAS Antireflective-Antistatic ııı.ARC Archive (dosya ad uzant s )ARCA Advanced RISC Computing ArchitectureARCnet Attached Resource Computer Network ııı.ARJ Compressed File (dosya ad uzant s )ARL Adjusted Ring LengthARLL Advanced Run Length LimitedARM Advanced RISC Machine (processor) +ARM Annotated Reference Manual +ARM Asynchronous Response ModeARMA Association of Records Managers and AdministratorsARP Address Resolution ProtocolARPANET Advanced Research Projects Agency NetworkARPL Adjust Requested Privilege LevelARS Activity Reporting System (Unisys)ARTA Apple Real Time ArchitectureARTIC A Real-Time Interface Coprocessor (IBM)ARTS Asynchronous Remote Takeover Server.ARTS Cultural Groups (Domain Name) (Internet)ARTT Asynchronous Remote Takeover TerminalARQ Automatic Repeat RequestARU Audio Response UnitAS Autonomous System (Internet)ASAI Adjunct Switch Application Interface (AT&T)ASAP As Soon As Possible +Automatic Switching And ProcessingAS3AP ANSI SQL Standard Scalable and PortableASB Advanced System Buffering (Intel)ııı.ASC ASCII text (dosya ad uzant s )ASCC Automatic Sequence Controlled Calculator (IBM)ASCII American Standard Code for Information InterchangeASD Application Structure Database (Microsoft) +Automatic Skip Driver (Microsoft)ASF Active/Advanced Streaming Format (Microsoft)ASIC Application-Specific Integrated CircuitASIT Advanced Security and Identification Technology ASL Adaptive Speed Levelling ASLM Apple Shared Library Manager (Macintosh)ııı.ASM Assembler Source Language (dosya ad uzant s )ASMP Asymmetric MultprocessingASN Abstract Syntax Notation + Autonomous System NumberASO Automated Systems OperationsASP Active Server Page (Microsoft) +ASP Application/Authorized Service Provider +ASP Aspect Oriented Programming +ASP Association of Shareware Professionals ııı.ASP Active Server Page (dosya ad uzant s )ASPI Advanced SCSI Programming Interface (Adaptec)ASPS Advanced Signal Processing SystemASP Address Space Register + Automatic Send-Receive +ASP Automatic Speech RecognitionAS/U Advanced Server for Unix ASYNC Asynchronous AS/400 Application System/400 (IBM)AT Advanced Technology + Attention ATAPI Advanced Technology Attachment Packet Interface ATC Advanced Transfer CacheATDP Attention Dial Pulse ATDT Attention Dial Tone ATE Automated Test Equipment ATG Advanced Technology Group ATH Attention Hang-Up (Modem command)ATL Active Template Library ATM Adobe Typeface Manager + Air Traffic Management +ATM Asynchronous Transfer Mode + Automated Teller Machine ATPS AppleTalk Printing Services (Macintosh)ATR Automatic Terminal Recognition ATS Administrative Terminal System +ATS Apple Terminal Services AT&T American Telephone and TelegraphATTN Attention ııATTRIB Attribute (bak n z. ATR)ATX Advanced Technology Extended (Intel)ııı.AU Audio (dosya ad uzant s )AUDIT Automated Data Input TerminalAUI Attachment/Autonomous Unit InterfaceAUP Acceptable Use Policy (Internet)AUTO AutomaticAUTOEXEC Automatic ExecutionAUX Auxiliary + (First Serial Port)AV Audio/Video + Audiovisual + Authenticity VerificationAVA Audio Visual Authoring (IBM)AVC Audio Visual Connection (IBM)AVD Alternating Voice and DataAVG AverageAVI Audio Visual Interleave (Microsoft)AVP Attribute-Value PairAVR Automatic Voice RecognitionAVT Applied Voice TechnologyAVVID Architecture for Voice Video & Integrated Data (Cisco)AWE Advanced Wave Effects (Creative Lab)AWG American Wire GageAWS Advanced Workstations and Systems (group) (IBM)AWT Abstract Window Toolkit (Java)AX Architecture Extended + Automatic TransmissionBACP Bandwidth Allocation Control ProtocolBAK Binary Adaptation Kit (Microsoft)ııı.BAK Backup (dosya ad uzant s )BAL Basic Assembly LanguageBAM Boyan Action ModuleBAPI Business Application Programming Interface (SAP)BARTS Bell Atlantic Regional Timesharing ııı.BAS Basic Language (dosya ad uzant s )BASH Bourne Again Shell (Unix)BASIC Beginner's All-purpose Symbolic Instruction CodeBASM Built-In AssemblerBAT Block Address Translation ııı.BAT Batch Processing (dosya ad uzant s )B2B Business-To-Business BBS Bulletin Board System B2C Business To ConsumerBCC Block Check Character BCC: Blind Carbon Copy BCD Binary Coded DecimalB-CHANNEL Bearer Channel BCL Base Class Libraries + Batch Command LanguageBCN BeaconBCNF Boyce-Codd Normal FormBCP Best Current Practice + Bulk Copy ProgramBCPL Basic Computer Programming LanguageBCR Byte Count RegisterBCS Bar Code SorterBDA Bios Data AreaBDC Backup Domain ControllerBDE Borland Database Engine (Borland)BDLS Bidirectional Loop SwitchingBDOS Basic Disk Operating SystemBDR Bus Device Request ııBDSL (Bak n z DSL)BE Below or Equal B2E Business To Employees BECN Backward Explicit Congestion Notification BEDO Burst Extended Data OutBEEP Blocks Extensible Exchange Protocol BEL Bell BELLCORE Bell Communications Research BER Basic Encoding Rules + Bit Error Rate BERT Bit Error Rate Test/Tester BF Bad Flag B/F Background/Foreground ııı.BFC Briefcase (dosya ad uzant s ) (Microsoft)BFF Binary File Format (IBM)BFT Binary File Transfer BFTP Batch FTP BGA Ball Grid Array BGCOLOR Background Color (HTML)BGE Branch if Greater or Equal ııı.BGI Borland Graphic Interface (dosya ad uzant s )BGP Border Gateway ProtocolBGT Branch if Greater ThanBHI Branch if HigherBHIS Branch if Higher or SameBI Binary Input ııı.BIB Bibliography (dosya ad uzant s )BiDi Bi-DirectionalBIFET Bipolar Field Effect TransistorBIFF Binary Interchange File FormatBIM Beginning of Information Marker ııı.BIN Binary (dosya ad uzant s )BINAC Binary Automatic ComputerBIND Berkeley Internet Name DomainBINHEX Binary HexadecimalBIOS Basic Input/Output SystemBIS Business Information SystemBISDN Broadband Integrated Services Digital NetworkBIST Built-In Self-Test ııBiSYNC (Bak n z BSYNC)BIT Binary DigitBITNET Because It's Time NetworkBITNIC Bitnet Network Information CenterBIU Bus Interface UnitBIW Business Intelligence Warehouse (SAP)BIX Byte Information Exchange (BBS).BIZ Businesses (Domain Name) (Internet)ııı.BK! Backup (dosya ad uzant s ) (WordPerfect)ııı.B~K Backup (dosya ad uzant s )BKSP BackspaceBL Backlit + Bit LineBLAST Bell Labs Layered Space Time +BLAST Blocked Asynchronous Transmission (protocol)ııı.BLD BASIC Bload Graphics (dosya ad uzant s )BLE Branch if Less or EqualBLER Block ErrorBLK BlockBLMC Buried Logic Macrocell BLOB Binary Large Object BLOS Branch if Lower Or SameBMI Branch if Minus BMIC BusMaster Interface Controller (Intel)BMP Basic Mapping Support + Basic Multilingual Plane +BMP Batch Message Processing Program ııı.BMP Bitmap (dosya ad uzant s ) (Paintbrush Format)ıBNC Bayonet Neill-Concelman (Mr. Neill-Concelman taraf ndan bulunan konnekt ör) +BNC British Naval ConnectorBNE Branch if Not EqualBNS Backbone Network ServiceBO Binary OutputBOA Basic Object AdapterBOB Break-out BoxBOC Basic Operator ConsoleBOF Beginning Of FileBOM Basic Online Memory (IBM) + Beginning Of MessageBOND Bandwidth On DemandBOOTP Bootstrap Protocol (Internet)BOPS Billion Operations Per Second BORPQU Borland Pro Quattro BORQU Borland Quattro BOS Basic Operating System BOT Beginning Of Table + Beginning of Tape + RobotBP Base Pointer BPB BIOS Parameter Block BPDU Bridge Protocol Data Unit BPDU Berkeley Packet Filter BPI Bits Per Inch BPL Branch if Plus BPP Bits Per Pixel BPR Business Process Re-engineering (Linux)BPS Bits Per Second + Bytes Per SecondBPSK Binary Phase Shift Keying BR Bad Register BRGC Binary Reflected Gray Code BRI Basic Rate Interface + Brain Response InterfaceBS BackspaceBSAM Basic Sequential Access MethodBSC Binary Synchronous CommunicationBSCS Bachelor of Science (Degree) in Computer ScienceBSD Berkely Software/Standard DistributionBSF Bit Scan ForwardBSI British Standards InstituteBSM Basic Storage Module (IBM)BSP Bulk Synchronous ParallelismBSR Bit Scan ReverseBS-RAM Burst Static RAMBSS Block Started by SymbolBST Binary Search TreeBSY BusyBSYNC Binary Synchronous Communications (protocol)BT Bit TestBTAM Basic Telecommunications Access Method (IBM)BTB Branch Target BufferBTC Bit Test and ComplementBTI Business Type IdentifierBTOA Binary To ASCIIBTP Batch Transfer ProgramBTR Bit Test and ResetBTS Bit Test and SetBU Branch UnitBUBL Bulletin Board for Libraries (Janet)BUF BufferBUS Broadcast and Unknown ServerBVH Base Video HandlerBWM Block-Write ModeB2X Binary To Hexadecimal (REXX)BYTE Binary Element StringC C Programming Language ııı.C C source code (dosya ad uzant s ) (C)CA Certification Authority + Collision AvoidanceCAB Compressed Application Binary (Microsoft)ııı.CAB Cabinet (dosya ad uzant s ) (Microsoft)CACLS Command-Line Access Control Lists (Microsoft)CAD Computer Aided Design CADD Computer Aided Design and Drafting CADE Client/server Application Development EnvironmentCAE Client Application Enabler + Common Applications CAE Environment + Computer Aided Engineering CAEN Chemically Assembled Electronic NanocomputerCAEX Computer Aided Exploration CAG Column Address GeneratorCAI Computer Aided InstructionCAIRN Collaborative Advanced Interagency Research NetworkCAL Client Access License +Computer Aided LearningCALS Computer-Aided Acquisition in Logistic SupportCAM Common Access Method + Computer Aided Manufacturing +CAM Contents Addressable MemoryCAN Cancel + Controller Area NetworkCANBUS Conroller Area Network BusCAP Carrierless Amplitude and Phase-Modulation +CAP Central Arbitration Point +CAP Communication Application Platform +CAP Competitive Access Provider +CAP Computer Aided Publishing ııı.CAP Capture (dosya ad uzant s )CAPE Concurrent Art-to-Product Environment CAPS Capitals (Upper Case Letters) +CAPS Cassette Programming SystemCAR Committed Access Rate (Cisco)CARL Colorado Alliance of Research Libraries (Internet)CART Computer-Assisted Radar Tomography CAS Column Address Select/Strobe +CAS Communications Application Specification +CAS Computer Aided Styling CASE Computer Aided Software Engineering CASL Crosstalk Application Scripting Language (DCA)CASS Computer Assisted Search ServiceCASSIS Classified and Search Support Information System CAT Computer Adaptive Test + Computer Aided Testing +CAT Computer Aided Tomography +CAT Computer Aided Transcription + Concatenate ııı.CAT Catalog (dosya ad uzant s )CATS Computer Assisted Training SystemCAT SCAN Computerized Axial Tomography ScanCATV Community Antenna Television (Cable TV)CAU Controlled Access UnitCAV Constant Angular VelocityCAVE Computer Automatic Virtual EnvironmentCBC Cipher Block ChainingCBCR Channel Byte Count RegisterCBDS Connectionless Broadband Data ServiceCBEMA Computer and Business Equipment Manufacturers AssociationCBGA Ceramic Ball Grid ArrayCBI Computer Based Instruction\InstrumentationCBL Computer Based Learning ııı.CBL COBOL source code (dosya ad uzant s )CBMS Computer-Based Mail SystemCBR Case Based Reasoning + Constant Bit RateCBT Computer Based TrainingCBW Convert Byte to WordCBX Computer-Controlled Branch ExchangeCC Cluster ControllerCC: Carbon CopyCCD Charged-Coupled DeviceCCFT Cold Cathode Fluorescent TubeCCI Common Client InterfaceCCIE Cisco Certified Internetwork Expert (Cisco)CCL Connection/Cursor Control LanguageCCM Corba Component ModeCCNA Cisco Certified Network AssociateCCNP Cisco Certified Network ProfessionalCCP Certified Computing Professional + Console Command ProcessorCCS Common Command Set + Common Communications Services +CCS Common Communications Support +CCS Continuous Composite Servo + Cryptographic Checksum ııı.CCS Color Calibration System (dosya ad uzant s )CCSD Cellular Circuit-Switched DataCCTLD Country Code Top-Level DomainCD Carrier Detect + Change Directory +CD Collision Detection + Color Display + Compact Disk C/D Control Data C2D Character To Decimal (REXX)CDA Compound Document Architecture (DEC)ııı.CDA CD Audio Track (dosya ad uzant s )CDBT Compact Disk-Based TrainingCDC Connected Device Configuration + Control Data Corporation CD_CHRDY Card Channel Ready (IBM)CD-DA Compact Disk -Digital AudioCDDI Copper Distributed Data InterfaceCDE Common Desktop Environment + Complex Data EntryCD-E Compact Disk - ErasableCDF Channel Definition Format ııı.CDF Comma Delimited Format (dosya ad uzant s )CDFS Compact Disc File System (Microsoft)CD+G Compact Disk plus GraphicsCD-I Compact Disk - InteractiveCDIA Certified Document Imaging ArchitectCDL Computer Design LanguageCDMA Code Division Multiple Access CDMF Commercial Data Masking Facility (IBM)CD-MO Compact Disk - Magneto Optical CDO Collaboration Data Objects (Microsoft)CDOS Concurrent Disk Operating System CDPD Cellular Digital Packet DataCDR Call Detail Record + Call Detail Recording +CDR Common Data Representation ııı.CDR CorelDRAW (dosya ad uzant s ) (Corel)CD-R Compact Disk - Recordable CD-RAM Cached RAM CD-RDx Compact Disk - Read Only Memory Data Exchange Standard CDRL Contract Data Requirements List CD-ROM Compact Disk - Read Only Memory CD-RTOS Compact Disk - Real Time Operating SystemCD-RW Compact Disk - Rewritable CDS Current Directory Structure CDSA Common Data Security Architecture ııCDSL (Bak n z DSL)ııı.CDT Corel Draw Template (dosya ad uzant s )CD-V Compact Disk - VideoCD-WO Compact Disk - Write Once ııı.CDX Compound Index (dosya ad uzant s ) (Fox Pro)CD-XA Compact Disk - Extended ArchitectureCE Cache Enable + Chip Enable + Collision Elimination +CE Convert EnableCEC Certified E-Commerce ConsultantCEG Continuous Edge GraphicsCEI Conducted Electromagnetic InterferenceCELP Card Edge Low Profile (socket) (Intel)CEMS Constituent Electronic Mail SystemCEOP Conditional End Of PageCER Canonical Encoding RulesCERN The European Particle Physics LaboratoryCERT Computer Emergency Response TeamCFB Cipher Feedback + Configurable Function Block ııı.CFG Configuration (dosya ad uzant s )CFML ColdFusion Markup LanguageCFR Computerized Facial RecognitionCFS Caching/Common File SystemCFV Call For Votes (Bitnet)CG Control GateCGA Color Graphics AdapterCGE Common Graphics EnvironmentCGI Computer Generated Images + Computer Graphics InterfaceCGI-BIN Common Gateway Interface - Binary ıııııı.CGM Computer Graphics Metafile (dosya ad uzant s ) + Graph (dosya ad uzant s ) (Lotus 1-2-3) (LDC)CGS Continuous-Grain SiliconCHAP Challenge-Handshake Authentication ProtocolCHAR CharacterCHAT Conversational Hypertext Access Technology (Internet)CHCK Channel CheckCHCP Change Code PageCHDIR Change DirectoryCHFN Change Finger (Unix)CHGRP Change Group CHIPS Clearinghouse Interbank Payments System ııı.CHK CHKDSK (dosya ad uzant s )CHKDSK Check Disk CHMOD Change Mode CHOWN Change OwnerCHP Chapter CHR CharacterCHRP Common Hardware Reference PlatformCHS Cylinder Head SectorCHTML Compressed HTMLCI Component InterfaceCIA Current Instruction AddressCIAC Computer Incident Advisory CapabilityCICS/VS Customer Information Control System / Virtual Storage (IBM)CID Charge-Injection Device + Configuration/Installation/DistributionCIDR Classless Inter-Domain RoutingCIF Common Interchange/Intermediate Format + Crystallographic Information FileCIFS Common Internet File SystemCIL Common Intermediate Language (ECMA)CIM Common Information Model + CompuServe Information Manager +Computer Integrated Manufacturing CIO Chief Information Officer CIOCS Communication Input/Output Control System CIP Command Interface Port + Common Indexing Protocol CIR Committed Information RateCIRC Circular Reference + Cross-Interleaved Reed-Solomon Code CIS Card Information Structure + Client Information System +CIS CompuServe Information Service +CIS Computer Information Systems + Contact Image Sensor +CIS Customer Information System CISC Complex Instruction Set Computing CIT Computer-Integrated Telephony CIVR Computer and Interactive Voice Response CIW Certified Internet WebmasterCIX Commercial Internet Exchange + Compulink Information Exchange CJLI Command Job Language Interpreter CKD Count Key Data (device)CLAR Channel Local Address RegisterCLASS Client Access to Systems and Services +CLASS Cooperative Library Agency for Systems and Services +CLASS Custom Local-Area Signaling ServicesCLC Clear Carry FlagCLD Clear Direction FlagCLDC Connected Limited Device ConfigurationCLEC Competitive Local Exchange CarrierCLI Call-Level Interface + Clear Interrupt Flag +CLI Client Library Interface + Command Line Interface +CLI Common Language Infrastructure/InterfaceCLIB C LibraryCLID Calling Line IdentificationCLIST Command ListCLK ClockCLNP Connectionless Network ProtocolCLOS Common Lisp Object SystemCLP Cell Loss Priority + Constraint Logic Programming ııı.CLP Clipboard (dosya ad uzant s ) (Microsoft)CLR Common Language Runtime (Microsoft)CLS Clear Screen + Common Language SupportCLSID Class IdentifierCLTP Connectionless Transport ProtocolCLTS Clear Task Switch FlagCLTV Constant Linear Time VelocityCLUI Command Line User InterfaceCLUT Color Look-Up TableCLV Constant Linear VelocityCM Centimeter + Control Mark + Corrective MaintenanceCMA Concert Multi-thread ArchitectureCMC Common Mail Calls + Common Messaging Calls +CMC Communication Management Configuration +CMC Complement Carry Flag +CMC Computer-Mediated Communication (Internet)CMD Circuit Mode Data ııı.CMD Command (dosya ad uzant s )CMF Creative Music FormatCNIDR Clearinghouse for Network Information and Discovery and Retrieval (Internet)CMIP Common Management Information ProtocolCMIS Common Management Information Services/System CML Chemical Markup Language + Conceptual Modelling Language +CML Computer Managed Learning + Current Mode LogicCMM Capability Maturity Model CMMS Computerized Maintenance Management Software CMMU Cache/Memory Management Unit (Motorola)CMOS Coats-Mellon Operational Specification +CMOS Complementary Metal-Oxide SemiconductorCMOV Conditional MoveCMP Compare + ComputerCMPS Compare Word StringCMS Call Management Server + Code Management System +CMS Compiler Monitor System + Conversation Monitor SystemCMTS Cable Modem Termination SystemCMVC Configuration Management Version Control (IBM)CMW Compartmented Mode WorkstationCMY Cyan-Magenta-Yellow (color model)CMYK Cyan-Magenta-Yellow-Black (color model)CNA Certified NetWare/Network AdministratorCNAPS Co-Processing Node Architecture for Parallel Systems CNC Computerized Numerical Control CNE Certified NetWare Engineer ııı.CNF Configuration (dosya ad uzant s )CNG Calling (tone)CNI Certified Novell Instructor (Novell)CNIDR Clearinghouse for Networked Information and Discovery and Retrieval (Internet)CNN Composite Network Node CNR Carrier to Noise Ratio + Communication Network Riser CNSS Core Nodal Switching Subsystem (Internet)ııı.CNT Contents (dosya ad uzant s )CNV Conventional (pertaining to Memory)CNVT Convert CNX Certified Network Expert CO Central Office + Command Output + Convert OutCOAST Card On A Stick (module)COAX Coaxial Cable COB Chip-On-Board ııı.COB COBOL source code (dosya ad uzant s )ııCOBOL Common Business-oriented Language (Bak n z HLL)ııı.COD Code List (dosya ad uzant s )CODASYL Conference on Data System Languages (Group thatdesigned COBOL)CODE Client-Server Open Development EnvironmentCODEC Coder/Decoder + Compression/DecompressionCOEM Commercial Original Equipment ManufacturerCOFF Common Object File Format (Unix)COGO Coordinate Geometry (Programming Language)COL Collision + Computer Oriented LanguageCOLD Computer Output to Laser DiskCOLL CollisionCOM Component Object Model (Microsoft) + Computer Output Microfilm ııı.COM Command (dosya ad uzant s ) + Commercial Business (Domain Name) (Internet)COM1 First serial Port (asynchronous port)COM2 Second serial PortCOM3 Third serial PortCOM4 Fourth serial PortCOMDEX Computer Dealers ExpositionCOMET Cornell Macintosh Terminal EmulatorCOMM CommunicationsCOMP CompareCOMP. Computers (USENET Newsgroup Category)COMSAT Communications Satellite CorporationCON Console (includes Keyboard and Screen)COND ConditionCONFIG ConfigurationCONS Connection-Oriented Network ServiceCONTONE Continuous Tone.COOP Cooperatives-Business (Domain Name) (Internet)COPICS Communications Oriented Production Information and Control System (IBM)COR Common Object RuntimeCORBA Common Object Request Broker ArchitectureCOREN Corporation for Research and Enterprise NetworkCOS Compatible Operating SystemCOSE Combined Office Standard Environment + Common Open Software/Systems Environment COSMIC Computer Software Management and Information Center (NASA)COSMOS Computer System for Mainframe OperationsCOSS Common Object Services SpecificationCOTS Commercial Off-The-Shelf (software)CP Copy Protected CPA Certified Public Accountant + Cost Per ActionCPAN Comprehensive Perl Archive Network CPC Constant Point Calculation + Cost Per Click CPCS Check Processing Control System (IBM)CPE Central Processing Element + Customer Premises Equipment +Customer Provided Equipment ııı.CPE Cover Page (dosya ad uzant s )。
WELCOMEBUILD APPS & EXTEND DREAM IT THEN BUILD IT ON DOMOTrevor Tingey Director of Engineering Domo, App PlatformChris JonesCo-founder, Chief Solutions GuruData JoNZDREAM IT, THEN BUILD IT ON DOMO.StoreA FOCUS ON SPEED TO VALUE Enterprise Applications Build on Domo's open cloud platform to create custom solutions as unique as your business.3Analytical Analytics-first approach throughcore Domo.1Fast Faster setup with a focus on need .2Scalable Auto-scaling and distribution of business logic.DOMO Platform Business NeedsThe code that’s the fastest to write, the code that’s the easiest to maintain, and the code that never breaks,is the code you don’t write.-Steve JobsTHE KEY TO FAST, SERVERLESS APP DEVELOPMENT Typical Application Development HardwareServers/Applications/ DistributionUser Management/Security File and DB StorageApplication Business Logic IaaS Infrastructure-as-a-Service YOU’RE NOW HEREPaaSPlatform-as-a-Service YOU WERE HEREDOMO’s Enterprise ApplicationsEnterprise Application Platform ApplicationServerless Development/PlatformCLOUD PLATFORM SIMPLICITYWith Domo’s server-less cloud platform,you save time and money with no needfor purchasing hardware, server setup,or storage costs.LASER FOCUSKeep your team focused on appdevelopment instead of infrastructure.Domo's architecture dynamically scalesto trillions of records and thousands ofusers.USER-FRIENDLYGet Domo’s out-of-the-box chartingengine with interactive charts topower all your app visualizations.BUILT-IN SCALEDomo’s server-less cloud platformdynamically scales to trillions ofrecords and thousands of users.BEST DISTRIBUTIONShare your app with your team withbuilt-in authentication and SSOintegration.SERVERLESS DEVELOPMENT: STORING DATA Creating, Saving, Retrieving and Querying objectsSERVERLESS DEVELOPMENT: CONSUMING SERVICESInterfacing with other services, i.e.•Messaging (email, sms)•Exporting engines, i.e. PDF•AI/Machine Learning EnginesSERVERLESS DEVELOPMENT: BEING SMART•Wrapping up multiple interactions & combining the results•Abstracting complicated business logic•Orchestrating workflow processesSERVERLESS DEVELOPMENT AT CLOUD SCALEUser Management Built-in Authentication / SSO IntegrationFile Storage Scalable and Version Controlled Data WarehouseIntegration with all your Domo Data DistributionShare your App with the World Domo Actions Executable Serverless FunctionsEnterprise ApplicationsAppDB Schema-less Application State PersistenceApps can allow your customers to fill out a form that is then saved to AppDB. That data can then be used directly in real-time from your app or synced back to Domo as a Domo DataSet forreporting, etc.USE CASE: FORMSData-driven maps, graphics and media-rich layouts with configurable, custom experiences are just a few examples of apps that are creating immersiveexperiences within Domo customer’s businesses.USE CASE: ADVANCED DATA VISUALIZATIONSMaking decisions on your data doesn’t have to stop with analysis. With an app you can take action not only inside of Domo, but also by calling out to 3rd party services to make adjustments toyour business systems.USE CASE: TAKE ACTIONAn app built on Domo is uniquely positioned to take advantage of all of your data. You can then use that data to make better decisions by implementing machine learning or enhancingmanual processes and automating E CASE: AUGMENTED INTELLIGENCEData JoNZ Solutions Built on DomoWHY WE BUILD ON DOMO WHY WE BUILD ON DOMOBest of breed transformational platformPlace where all the data livesPlatform that focuses on user empowermentComplete suite of managed services and API building blocksFlexibility -leverage the tools and technologies we use to build Rapid deployment/multi-channel capabilitiesConnected across the enterpriseSolution opportunities –forward thinking companyDOMO DIGITAL TRANSFORMATION PLATFORM 1000+ no-code API connectorsAdrenaline powered warehouse queries scale to any size dataEnterprise grade security and governanceNo-code and low code ETL options empower business usersHundreds of insights and visualization’sEasy to use storyboard and interface builder toolsIntegrated collaborationApp embeddingBest of breed mobile app experienceApp Store with 1000+ custom insights, applications, connectors, and dataA surface that sits over the enterprise and connects all pointsCREATE TRANSFORMATION AT SCALE Streamlining workflows tied to Excel,PDF, webforms, legacy systems, paperHaving analytics at the point of dataentry and decision makingConnecting workflows with analytics, communications, and business systemsFinding robust solutions that don’t heavily depend on IT staff/power usersUser ManagementBuilt-in Authentication /SSO IntegrationFile StorageScalable and VersionControlled Data WarehouseIntegration with all yourDomo DataDistributionShare your App withthe WorldDomo ActionsExecutable ServerlessFunctions AppDBSchema-less ApplicationState PersistenceDOMO APP SOLUTION –LARGE TECHNOLOGY VENDOR ITEM CODING PROCESSSelling a mix of OEM and third-party equipment, software, services –Hundreds of products and services with thousands of variations–Daily updates and new itemsStruggling with an Excel form and Outlook-based approval process for getting product information into Oracle-based ERP system and price book –Slow, error prone–Products cannot be sold until in system–No ability to escalate, see bottlenecks–Wanted an archive and ability to generate metricsHaving IT dept. develop a custom solution was cost and time prohibitiveLEGACY FORM6. Upload to Oracle 3. Review, Edit,Comment, ApproveItem Specialists2. Data EntryOEM & Third-party suppliersNew product (HW, SW,service)Updates to existingproducts1. Product InfoManualrequest foritem code4. CodeRequestManualentry5. CodeEntryCUSTOM APP D-FLOW TRANSFORMED PROCESS Jsd;ü•See relevant forms, fields •Bulk uploading, versions •Strong edit tracking, archive•Parallel, automated process •Rules/role-based workflow •Reminders, status data 2. Data Entry1. Product Info3. Review, Edit, Comment,Approve4.Automated Entry5. Real time analytics, archiveForms Designer WorkflowEngineWebhooksOutputBuilder Domo PlatformUserPortal d-FlowBringing Apps to Life -DEMOKEY TAKEAWAYSDomo’s managed services platform is hereEmbracing a modern architecture of micro-services and micro-applications Connecting & building on your data and analytics.Save time and money building on where everything comes togetherLess code more solutionsDream it and you can build itGet started at the Domo developer portal today!D-FLOWLIMITLESS WORKFLOW APPLICATIONSD-FLOW CODELESS APPLICATION BUILDERD-TOOLSWAREHOUSED-OUTLOOKFOR MORE INFORMATION VISITTHANK YOU。
Aurora 5 is the automated file-based QC tool that integrates into your production workflow and consistently delivers dependable results. We focus on eliminating false positives and utilize a high degree of correlation to human audio and visual perception to ensure that our test reports highlight just the key issues, not hundreds of irrelevant non-issues.The unique Aurora 5 architecture and advanced use of both CPUs and GPUs ensures concurrent analysis of more file types at faster speeds than alternative solutions. Aurora is available with configurable CPU allocation and 3 levels of product for different grades of hardware utilization, providing an upgrade path for increased performance in the future.Constant feedback from customers, integrators and partners helped design the new Aurora 5 user interface, making configuration, operation and review easy to learn, quick to adopt and practical across a variety of ingest, production, and distribution environments. Whether you are looking to apply metadata gate-keeping to keep your ingest clean, detecting visual and audible artifacts, identifying common editing errors, or testing for broadcast and distribution constraints, Aurora 5 will make a positive difference across your operation. Aurora 5 StandardOur entry level Aurora 5 Standard file-based QC platform enables up to 2x CPU cores to be allocated to each Aurora Verification Unit (VU). The resulting performance enables throughput of SD files up to 5x real time and HD files up to real time.Aurora 5 ProfessionalOur flagship Aurora 5 Professional QC platform can utilize up to 8x CPU cores per Verification Unit and provides support from up to 2 GPUs.A full hardware configured Aurora 5 can analyze SD files up to13x real time, HD files up to 4x real time, and 4K files up to real time.Aurora 5 Professional PlusThe VU Plus Plug-in for Aurora 5 Professional enables up to 16 CPU cores to be dedicatedto each Verification Unit and will access high numbers of GPU cores for accelerated testing. The analysis speed is up to about 50% faster than Aurora 5 Professional.Aurora 5 is the new generationof automated file-based QC product that represents a significant leap forward for file-based quality control, moving it significantly ahead of other products in its class in a number of critical areas....• Unparalleled, Scalable Speed• Faster than Real-time 4K Analysis• Enhanced Adaptive Bit Rate Support• 8x Faster Manual Review• Configurable Automated Workflows• QC Report AnalyticsAurora 5 file-based QC 103/2015 Copyright © Tektronix. All rights reserved. 2NW-60054-0Guaranteed QC CapacityAurora 5 guarantees QC capacity by reserving CPU andCUDA GPU capacity so that the system does not to overload itself. Aurora 5 doesn’t slow down as the system loads up, with the first file analyzed just as fast as that same file with hundreds of other concurrent jobs in progress.Broad Codec and Wrapper SupportAs a result of constant participation in key standardscommittees (EBU, SMPTE, AMWA, FIMS) Aurora 5 always has the latest codes and wrappers with new updates arriving often. This means that almost any file you can make, Aurora can test.New User InterfaceAurora 5 has a totally updated user interface with a consistent information layout, with the ability to modify the theme/colors and display different language scripts in both the user interface and QC reports. Current support includes English, Japanese and Chinese Simplified.Easy-to-Use QC ReportsAurora 5 job reports are a single easy-to-read page displaying all of the metadata in one location followed by an interactive error summary including frame-accurate access through Hydra Player. The report is consistent in look and content whether using a web browser and PDF .Fast Access Help FilesAurora 5 provides Help files for every test accessed with a click from test templates or reports, providing anexplanation of each individual test, how the test is performed, recommended correction(s), and where to in the workflow may be the best location to perform the fix.Email NotificationAurora 5 email notifications ensure that regardless of the receiving email device there is appropriate informationprovided for decision making or further action. Emails contain summary info and a HTML link to the full QC Report, and optionally the PDF version.8x Faster Manual ReviewAurora 5 is designed to dramatically reduce human review of QC results by 8 times compared to linear processes. By tightly integrating the Tektronix Hydra frame accurate player, instant access and visual aids simplify the tasks and increase efficiency.Minimal False PositivesAurora 5 algorithms are designed to reduce false positives, ensuring an accurate QC report. Tektronix operates a unique machine learning loop that ensures that continuous accuracy improvements of test algorithms based on customer supplied files with known QC artifacts.Video Essence TestsAurora 5 video essence tests include Macro block Noise, Up-conversion, Comb Artifacts, Field Order Swaps, Tape/Digital Hits, Perceptual & Film Artifacts, Black/Freeze Frames, Letterboxing/Pillarboxing, Color Bars, PSE/Flash Detection (Harding FPA), and Cadence Change.Audio Essence TestsAurora 5 audio essence tests include Silence, Drop-outs, Peaks (dBTP , PPM, dBFS), Average Levels (R128, ATSC,ARIB), Clipping, Snaps/Clicks/Pops, Test Tones, Phase Swaps and Hiss/Hum. Aurora 5 also applies a user-defined Audio Service Map for processing AES wrapped tracks or when mono channel audio essence tracks are tested together.Metadata TestsAurora 5 metatadata tests include Container Syntax, Video Essence Syntax, Dolby Audio Syntax, Dolby E Guard Band Alignment, Caption Syntax, Container Essence Contents, Cross-Check Container-Essence, Rude Word Detection in Text, Start Time code, Time code Discontinuity, Video Resolution and Run-times.Distribution Constraint TestsAurora 5 has pre-defined distribution tests with the distribution constraints pre-set. This includes test plans for CableLabs VOD, iTunes, Netflix, DVB, ISDB-T / TB, ATSC, XDCAM / RDD-9, AS-02, DPP / AS-11, DVD and Blu-ray.Automated WorkflowsAurora 5 includes tools for implementing end-to-end workflows, including Smart Test Plans for automatically applying test plans to your incoming files, and a Decision Engine that enables automated post QC test file movement and corrective actions.Systems IntegrationAurora QC includes an easy to use SOAP API that leading vendors have used to seamlessly integrate Aurora QC into their solution workflows. Major partners include Amberfin, Aspera, Astec, Imagine Communications, Telestream, iBroadcast, DataMiner, and FileCatalyst.Aurora 5 is the new generation of file-based QC solution from Tektronix fully implemented in a 64-bit architecture, enabling broad use of available CPU processing power and GPU acceleration to deliver unrivaled QC analysis speed and accuracy, perform up to 8 times faster manual review,and automate corrective actions.Unrivaled, Scalable Speed Aurora 5 is the first file-based QC product to have the ability to dynamically allocate threads across a user-specified number of CPU cores, enabling performance and scalability in high densityvirtual and blade environments. Using high CPU allocation or a lower CPU allocation combined with Aurora 5’s unique GPU accelerated processing capability, QC throughput approaching wireline limits can be achieved, rather than being constrained by a legacy decoding and testing architecture.QC Report AnalyticsAurora 5 is the first file-based QC product that has an optional QC Analytics plug-in to enable media organizations to analyze results of multiple QC Reports and search across reports for specific criteria. Aurora 5 can identify QC artifact trends and compare results from QC done at different workflow stages, pipeline issues and trends or identify vendors for KPI and SLA documentation.4K Production WorkflowAurora 5 can test and play back IMF , the master format agreed on by major motion picture studios. Aurora 5 has the ability to QC and play back complex Composition Play Lists (CPL) and is capable of testing 4K at real-time or faster in any other container or codec, including JPEG 2000, DPX, HEVC, H.264/AVC/AVCI/XAVC*. (* optionalAdvanced Codec pack is required for access to Canopus, HEVC and JPEG 2000 decoders.)Enhanced Adaptive Bit Rate SupportAurora 5 is the only file-based QC solution on the market capable of both testing and playing back HLS/HSS/HDS/DASH and CableLabs intermediate ABR playlist file sets. Aurora 5 includes tests that have been specifically designed to catch the most common causes of adaptive bit rate streaming problems.Aurora 5 Automated File-based QC with manual review player and workflowAs of April 2015 Aurora 5 supports the following file formats. Please contact Tektronix for latest listing.Supported FormatsASEAN / Australasia (65) 6356 3900Belgium 00800 2255 4835*Central East Europe and the Baltics +41 52 675 3777 Finland +41 52 675 3777Hong Kong 400 820 5835Japan 81 (3) 6714 3010Middle East, Asia, and North Africa +41 52 675 3777 People’s Republic of China 400 820 5835Republic of Korea 001 800 8255 2835Spain 00800 2255 4835*Taiwan 886 (2) 2722 9622Austria 00800 2255 4835*Brazil +55 (11) 3759 7627Central Europe & Greece +41 52 675 3777 France 00800 2255 4835*India 000 800 650 1835Luxembourg +41 52 675 3777The Netherlands 00800 2255 4835*Poland +41 52 675 3777Russia & CIS +7 (495) 6647564Sweden 00800 2255 4835*United Kingdom & Ireland 00800 2255 4835*Balkans, Israel, South Africa and other ISE Countries +41 52 675 3777 Canada 1 800 833 9200Denmark +45 80 88 1401Germany 00800 2255 4835*Italy 00800 2255 4835*Mexico, Central/South America & Caribbean 52 (55) 56 04 50 90 Norway 800 16098Portugal 80 08 12370South Africa +41 52 675 3777Switzerland 00800 2255 4835*USA 1 800 833 9200* European toll-free number. If not accessible, call: +41 52 675 3777For Further Information. Tektronix maintains a comprehensive, constantly expanding collection of application notes, technical briefs and other resources to help engineers working on the cutting edge of technology. Please visit .Copyright © Tektronix, Inc. All rights reserved. Tektronix products are covered by U.S. and foreign patents, issued and pending. Information in this publication supersedes that in all previously published material. Specification and price change privileges reserved. TEKTRONIX and TEK are registered trademarks of Tektronix, Inc. All other trade names referenced are the service marks, trademarks, orregistered trademarks of their respective 2NW-60054-010 Mar 2015Audio CodecsPCM Audio (WAV/AES/BWF), Dolby Digital (AC-3), DD+ (EAC-3), Dolby TrueHD (MLP), Dolby E, AAC, HE-AAC, WMA Standard/Pro, MPEG-2 (L1,2,3), MPEG-1.Captions/Subtitles/TextLine 21, CEA-608, CEA-708, Timed Text / DFXP , STL, SRT, SCC.Ordering informationPlease contact your Tektronix Sales Representative to understand how Aurora 5 may be customized for your specific workflow needs and content volume.Tektronix is registered to ISO 9001 and ISO 14001 by SRI Quality System Registrar..Container WrappersMXF (All OP , including AMWA defined AS, RDD-9, P2, SxS), Transport Stream, Elementary Stream, Program Stream/VOB, AVI, WMV/ASF , QuickTime/MOV , GXF , MP4, 3GPP , LXF , R3D, DPX, DXW, HLS, DASH, Smooth Streaming, IMF , DCP (unencrypted).Video CodecsH.264 (AVC/AVC-Intra/MVC 3D), MPEG-2 (including XDCAM, IMX and D-10), ProRes, JPEG 2000,DNxHD (VC-3), Cineform (VC-5), VC-1 (and WMV), DV/DVCPro, Flash VP-6/7, RAW (Huffman, YUV , RGB, Blackmagic), EXR, DPX, Canopus, H.265 (HEVC).。
The Dell EMC Networking S4048T-ON switch is the industry’s latest data center networking solution, empowering organizations to deploy modern workloads and applications designed for the open networking era. Businesses who have made the transition away from monolithic proprietary mainframe systems to industry standard server platforms can now enjoy even greater benefits from Dell EMC open networking platforms. By using industry-leading hardware and a choice of leading network operating systems to simplify data center fabric orchestration and automation, organizations can tailor their network to their unique requirements and accelerate innovation.These new offerings provide the needed flexibility to transform data centers. High-capacity network fabrics are cost-effective and easy to deploy, providing a clear path to the software-defined data center of the future with no vendor lock-in.The S4048T-ON supports the open source Open Network Install Environment (ONIE) for zero-touch installation of alternate network operating systems, including feature rich Dell Networking OS.High density 1/10G BASE-T switchThe Dell EMC Networking S-Series S4048T-ON is a high-density100M/1G/10G/40GbE top-of-rack (ToR) switch purpose-builtfor applications in high-performance data center and computing environments. Leveraging a non-blocking switching architecture, theS4048T-ON delivers line-rate L2 and L3 forwarding capacity within a conservative power budget. The compact S4048T-ON design provides industry-leading density of 48 dual-speed 1/10G BASE-T (RJ45) ports, as well as six 40GbE QSFP+ up-links to conserve valuable rack space and simplify the migration to 40Gbps in the data center core. Each40GbE QSFP+ up-link can also support four 10GbE (SFP+) ports with a breakout cable. In addition, the S4048T-ON incorporates multiple architectural features that optimize data center network flexibility, efficiency and availability, including I/O panel to PSU airflow or PSU to I/O panel airflow for hot/cold aisle environments, and redundant, hot-swappable power supplies and fans. S4048T-ON supports feature-rich Dell Networking OS, VLT, network virtualization features such as VRF-lite, VXLAN Gateway and support for Dell Embedded Open Automation Framework.• The S4048T-ON is the only switch in the industry that supports traditional network-centric virtualization (VRF) and hypervisorcentric virtualization (VXLAN). The switch fully supports L2 VX-• The S4048T-ON also supports Dell EMC Networking’s Embedded Open Automation Framework, which provides enhanced network automation and virtualization capabilities for virtual data centerenvironments.• The Open Automation Framework comprises a suite of interre-lated network management tools that can be used together orindependently to provide a network that is flexible, available andmanageable while helping to reduce operational expenses.Key applicationsDynamic data centers ready to make the transition to software-defined environments• High-density 10Gbase-T ToR server access in high-performance data center environments• Lossless iSCSI storage deployments that can benefit from innovative iSCSI & DCB optimizations that are unique only to Dell NetworkingswitchesWhen running the Dell Networking OS9, Active Fabric™ implementation for large deployments in conjunction with the Dell EMC Z-Series, creating a flat, two-tier, nonblocking 10/40GbE data center network design:• High-performance SDN/OpenFlow 1.3 enabled with ability to inter-operate with industry standard OpenFlow controllers• As a high speed VXLAN Layer 2 Gateway that connects thehypervisor based ovelray networks with nonvirtualized infrastructure Key features - general• 48 dual-speed 1/10GbE (SFP+) ports and six 40GbE (QSFP+)uplinks (totaling 72 10GbE ports with breakout cables) with OSsupport• 1.44Tbps (full-duplex) non-blocking switching fabric delivers line-rateperformance under full load with sub 600ns latency• I/O panel to PSU airflow or PSU to I/O panel airflow• Supports the open source ONIE for zero-touch• installation of alternate network operating systems• Redundant, hot-swappable power supplies and fansDELL EMC NETWORKING S4048T-ON SWITCHEnergy-efficient 10GBASE-T top-of-rack switch optimized for data center efficiencyKey features with Dell EMC Networking OS9Scalable L2 and L3 Ethernet switching with QoS and a full complement of standards-based IPv4 and IPv6 features, including OSPF, BGP and PBR (Policy Based Routing) support• Scalable L2 and L3 Ethernet switching with QoS and a full complement of standards-based IPv4 and IPv6 features, including OSPF, BGP andPBR (Policy Based Routing) support• VRF-lite enables sharing of networking infrastructure and provides L3traffic isolation across tenants• Increase VM Mobility region by stretching L2 VLAN within or across two DCs with unique VLT capabilities like Routed VL T, VLT Proxy Gateway • VXLAN gateway functionality support for bridging the nonvirtualizedand the virtualized overlay networks with line rate performance.• Embedded Open Automation Framework adding automatedconfiguration and provisioning capabilities to simplify the management of network environments. Supports Puppet agent for DevOps• Modular Dell Networking OS software delivers inherent stability as well as enhanced monitoring and serviceability functions.• Enhanced mirroring capabilities including 1:4 local mirroring,• Remote Port Mirroring (RPM), and Encapsulated Remote PortMirroring (ERPM). Rate shaping combined with flow based mirroringenables the user to analyze fine grained flows• Jumbo frame support for large data transfers• 128 link aggregation groups with up to 16 members per group, usingenhanced hashing• Converged network support for DCB, with priority flow control(802.1Qbb), ETS (802.1Qaz), DCBx and iSCSI TLV• S4048T-ON supports RoCE and Routable RoCE to enable convergence of compute and storage on Active FabricUser port stacking support for up to six units and unique mixed mode stacking that allows stacking of S4048-ON with S4048T-ON to providecombination of 10G SFP+ and RJ45 ports in a stack.Physical48 fixed 10GBase-T ports supporting 100M/1G/10G speeds6 fixed 40 Gigabit Ethernet QSFP+ ports1 RJ45 console/management port with RS232signaling1 USB 2.0 type A to support mass storage device1 Micro-USB 2.0 type B Serial Console Port1 8 GB SSD ModuleSize: 1RU, 1.71 x 17.09 x 18.11”(4.35 x 43.4 x 46 cm (H x W x D)Weight: 23 lbs (10.43kg)ISO 7779 A-weighted sound pressure level: 65 dB at 77°F (25°C)Power supply: 100–240V AC 50/60HzMax. thermal output: 1568 BTU/hMax. current draw per system:4.6 A at 460W/100VAC,2.3 A at 460W/200VACMax. power consumption: 460 WattsT ypical power consumption: 338 WattsMax. operating specifications:Operating temperature: 32°F to 113°F (0°C to45°C)Operating humidity: 5 to 90% (RH), non-condensing Max. non-operating specifications:Storage temperature: –40°F to 158°F (–40°C to70°C)Storage humidity: 5 to 95% (RH), non-condensingRedundancyHot swappable redundant powerHot swappable redundant fansPerformance GeneralSwitch fabric capacity:1.44Tbps (full-duplex)720Gbps (half-duplex)Forwarding Capacity: 1080 MppsLatency: 2.8 usPacket buffer memory: 16MBCPU memory: 4GBOS9 Performance:MAC addresses: 160KARP table 128KIPv4 routes: 128KIPv6 hosts: 64KIPv6 routes: 64KMulticast routes: 8KLink aggregation: 16 links per group, 128 groupsLayer 2 VLANs: 4KMSTP: 64 instancesVRF-Lite: 511 instancesLAG load balancing: Based on layer 2, IPv4 or IPv6headers Latency: Sub 3usQOS data queues: 8QOS control queues: 12Ingress ACL: 16KEgress ACL: 1KQoS: Default 3K entries scalable to 12KIEEE compliance with Dell Networking OS9802.1AB LLDP802.1D Bridging, STP802.1p L2 Prioritization802.1Q VLAN T agging, Double VLAN T agging,GVRP802.1Qbb PFC802.1Qaz ETS802.1s MSTP802.1w RSTP802.1X Network Access Control802.3ab Gigabit Ethernet (1000BASE-T)802.3ac Frame Extensions for VLAN T agging802.3ad Link Aggregation with LACP802.3ae 10 Gigabit Ethernet (10GBase-X) withQSA802.3ba 40 Gigabit Ethernet (40GBase-SR4,40GBase-CR4, 40GBase-LR4) on opticalports802.3u Fast Ethernet (100Base-TX)802.3x Flow Control802.3z Gigabit Ethernet (1000Base-X) with QSA 802.3az Energy Efficient EthernetANSI/TIA-1057 LLDP-MEDForce10 PVST+Max MTU 9216 bytesRFC and I-D compliance with Dell Networking OS9General Internet protocols768 UDP793 TCP854 T elnet959 FTPGeneral IPv4 protocols791 IPv4792 ICMP826 ARP1027 Proxy ARP1035 DNS (client)1042 Ethernet Transmission1305 NTPv31519 CIDR1542 BOOTP (relay)1812 Requirements for IPv4 Routers1918 Address Allocation for Private Internets 2474 Diffserv Field in IPv4 and Ipv6 Headers 2596 Assured Forwarding PHB Group3164 BSD Syslog3195 Reliable Delivery for Syslog3246 Expedited Assured Forwarding4364 VRF-lite (IPv4 VRF with OSPF, BGP,IS-IS and V4 multicast)5798 VRRPGeneral IPv6 protocols1981 Path MTU Discovery Features2460 Internet Protocol, Version 6 (IPv6)Specification2464 Transmission of IPv6 Packets overEthernet Networks2711 IPv6 Router Alert Option4007 IPv6 Scoped Address Architecture4213 Basic Transition Mechanisms for IPv6Hosts and Routers4291 IPv6 Addressing Architecture4443 ICMP for IPv64861 Neighbor Discovery for IPv64862 IPv6 Stateless Address Autoconfiguration 5095 Deprecation of T ype 0 Routing Headers in IPv6IPv6 Management support (telnet, FTP, TACACS, RADIUS, SSH, NTP)VRF-Lite (IPv6 VRF with OSPFv3, BGPv6, IS-IS) RIP1058 RIPv1 2453 RIPv2OSPF (v2/v3)1587 NSSA 4552 Authentication/2154 OSPF Digital Signatures Confidentiality for 2328 OSPFv2 OSPFv32370 Opaque LSA 5340 OSPF for IPv6IS-IS1142 Base IS-IS Protocol1195 IPv4 Routing5301 Dynamic hostname exchangemechanism for IS-IS5302 Domain-wide prefix distribution withtwo-level IS-IS5303 3-way handshake for IS-IS pt-to-ptadjacencies5304 IS-IS MD5 Authentication5306 Restart signaling for IS-IS5308 IS-IS for IPv65309 IS-IS point to point operation over LANdraft-isis-igp-p2p-over-lan-06draft-kaplan-isis-ext-eth-02BGP1997 Communities2385 MD52545 BGP-4 Multiprotocol Extensions for IPv6Inter-Domain Routing2439 Route Flap Damping2796 Route Reflection2842 Capabilities2858 Multiprotocol Extensions2918 Route Refresh3065 Confederations4360 Extended Communities4893 4-byte ASN5396 4-byte ASN representationsdraft-ietf-idr-bgp4-20 BGPv4draft-michaelson-4byte-as-representation-054-byte ASN Representation (partial)draft-ietf-idr-add-paths-04.txt ADD PATHMulticast1112 IGMPv12236 IGMPv23376 IGMPv3MSDP, PIM-SM, PIM-SSMSecurity2404 The Use of HMACSHA- 1-96 within ESPand AH2865 RADIUS3162 Radius and IPv63579 Radius support for EAP3580 802.1X with RADIUS3768 EAP3826 AES Cipher Algorithm in the SNMP UserBase Security Model4250, 4251, 4252, 4253, 4254 SSHv24301 Security Architecture for IPSec4302 IPSec Authentication Header4303 ESP Protocol4807 IPsecv Security Policy DB MIBdraft-ietf-pim-sm-v2-new-05 PIM-SMwData center bridging802.1Qbb Priority-Based Flow Control802.1Qaz Enhanced Transmission Selection (ETS)Data Center Bridging eXchange (DCBx)DCBx Application TLV (iSCSI, FCoE)Network management1155 SMIv11157 SNMPv11212 Concise MIB Definitions1215 SNMP Traps1493 Bridges MIB1850 OSPFv2 MIB1901 Community-Based SNMPv22011 IP MIB2096 IP Forwarding T able MIB2578 SMIv22579 T extual Conventions for SMIv22580 Conformance Statements for SMIv22618 RADIUS Authentication MIB2665 Ethernet-Like Interfaces MIB2674 Extended Bridge MIB2787 VRRP MIB2819 RMON MIB (groups 1, 2, 3, 9)2863 Interfaces MIB3273 RMON High Capacity MIB3410 SNMPv33411 SNMPv3 Management Framework3412 Message Processing and Dispatching forthe Simple Network ManagementProtocol (SNMP)3413 SNMP Applications3414 User-based Security Model (USM) forSNMPv33415 VACM for SNMP3416 SNMPv23417 Transport mappings for SNMP3418 SNMP MIB3434 RMON High Capacity Alarm MIB3584 Coexistance between SNMP v1, v2 andv34022 IP MIB4087 IP Tunnel MIB4113 UDP MIB4133 Entity MIB4292 MIB for IP4293 MIB for IPv6 T extual Conventions4502 RMONv2 (groups 1,2,3,9)5060 PIM MIBANSI/TIA-1057 LLDP-MED MIBDell_ITA.Rev_1_1 MIBdraft-grant-tacacs-02 TACACS+draft-ietf-idr-bgp4-mib-06 BGP MIBv1IEEE 802.1AB LLDP MIBIEEE 802.1AB LLDP DOT1 MIBIEEE 802.1AB LLDP DOT3 MIB sFlowv5 sFlowv5 MIB (version 1.3)DELL-NETWORKING-SMIDELL-NETWORKING-TCDELL-NETWORKING-CHASSIS-MIBDELL-NETWORKING-PRODUCTS-MIBDELL-NETWORKING-SYSTEM-COMPONENT-MIBDELL-NETWORKING-TRAP-EVENT-MIBDELL-NETWORKING-COPY-CONFIG-MIBDELL-NETWORKING-IF-EXTENSION-MIBDELL-NETWORKING-FIB-MIBIT Lifecycle Services for NetworkingExperts, insights and easeOur highly trained experts, withinnovative tools and proven processes, help you transform your IT investments into strategic advantages.Plan & Design Let us analyze yourmultivendor environment and deliver a comprehensive report and action plan to build upon the existing network and improve performance.Deploy & IntegrateGet new wired or wireless network technology installed and configured with ProDeploy. Reduce costs, save time, and get up and running cateEnsure your staff builds the right skills for long-termsuccess. Get certified on Dell EMC Networking technology and learn how to increase performance and optimize infrastructure.Manage & SupportGain access to technical experts and quickly resolve multivendor networking challenges with ProSupport. Spend less time resolving network issues and more time innovating.OptimizeMaximize performance for dynamic IT environments with Dell EMC Optimize. Benefit from in-depth predictive analysis, remote monitoring and a dedicated systems analyst for your network.RetireWe can help you resell or retire excess hardware while meeting local regulatory guidelines and acting in an environmentally responsible way.Learn more at/lifecycleservicesLearn more at /NetworkingDELL-NETWORKING-FPSTATS-MIBDELL-NETWORKING-LINK-AGGREGATION-MIB DELL-NETWORKING-MSTP-MIB DELL-NETWORKING-BGP4-V2-MIB DELL-NETWORKING-ISIS-MIBDELL-NETWORKING-FIPSNOOPING-MIBDELL-NETWORKING-VIRTUAL-LINK-TRUNK-MIB DELL-NETWORKING-DCB-MIBDELL-NETWORKING-OPENFLOW-MIB DELL-NETWORKING-BMP-MIBDELL-NETWORKING-BPSTATS-MIBRegulatory compliance SafetyCUS UL 60950-1, Second Edition CSA 60950-1-03, Second Edition EN 60950-1, Second EditionIEC 60950-1, Second Edition Including All National Deviations and Group Differences EN 60825-1, 1st EditionEN 60825-1 Safety of Laser Products Part 1:Equipment Classification Requirements and User’s GuideEN 60825-2 Safety of Laser Products Part 2: Safety of Optical Fibre Communication Systems FDA Regulation 21 CFR 1040.10 and 1040.11EmissionsInternational: CISPR 22, Class AAustralia/New Zealand: AS/NZS CISPR 22: 2009, Class ACanada: ICES-003:2016 Issue 6, Class AEurope: EN 55022: 2010+AC:2011 / CISPR 22: 2008, Class AJapan: VCCI V-3/2014.04, Class A & V4/2012.04USA: FCC CFR 47 Part 15, Subpart B:2009, Class A RoHSAll S-Series components are EU RoHS compliant.CertificationsJapan: VCCI V3/2009 Class AUSA: FCC CFR 47 Part 15, Subpart B:2009, Class A Available with US Trade Agreements Act (TAA) complianceUSGv6 Host and Router Certified on Dell Networking OS 9.5 and greater IPv6 Ready for both Host and RouterUCR DoD APL (core and distribution ALSAN switch ImmunityEN 300 386 V1.6.1 (2012-09) EMC for Network Equipment\EN 55022, Class AEN 55024: 2010 / CISPR 24: 2010EN 61000-3-2: Harmonic Current Emissions EN 61000-3-3: Voltage Fluctuations and Flicker EN 61000-4-2: ESDEN 61000-4-3: Radiated Immunity EN 61000-4-4: EFT EN 61000-4-5: SurgeEN 61000-4-6: Low Frequency Conducted Immunity。
DATA S H E E T Thunder CFWConsolidated Firewall, CGN, ADC, VPN and Secure Web Gatewayperformance data center firewall includes a Layer 4 firewall with integrated DDoS protection and server load-balancing to protect data center assets from the inside out.enterprises and service providers. In mobile networks, it can be deployed as a security gateway to enable secure backhaul between the RAN nodes and the core network.The policy-based IPsec VPN also enables high-capacity client-to-site VPN deployment to support remote access clients in enterprise network.in from access networks and roaming partners to support uninterrupted operations. Thunder CFW can secure its own resources, such as NAT IP pools, to ensure that operational functions are not compromised.Gi/SGi LAN PROTECTIONEvolvedPacket Core (EPC)Radio Access NetworkMobile & IoT DevicesA10 Gi/SGi Firewall for the GiLANIn this scenario, a mobile network operator deploys the Gi/SGi firewall to secure communication between the EPC and the internet to protect the mobile coreinfrastructure. Integrated carrier-grade NAT enables carriers to manage communication with both IPv4 and IPv6 address protocols. Built-in DDoS protection safeguards the NAT IP pools to avoid service interruption. A10 Harmony ® Controller provides centralized management and analytics.business intelligence, tighter security controls, enhanced lawenforcement agency compliance and service monetization.Agile Management and Array Analytics-driven DashboardGain application and network servicesvisibility with the A10Harmony Controllerfor Thunder CFW. Centrally configureandmanage policies across services in a multi-cloud environment. Get real-time actionableinsights on firewall performance, criticalCGN services such as mapping distribution,NAT IP pool utilization and more, andapplication visibility including applicationdistribution by category, bytes consumed byapplication category.spam and phishing sources. The A10 URL Classification categorizes more than 460 million domains and 13 billion URLs into 83 categories to block undesirable sites and shield users from threats.Block Known Web ThreatsIdentify and block traffic going out to and coming in from known bad IP addresses on the internet with threat intelligence feeds.Prevent Data ExfiltrationIntegrate with third-party Data Loss Prevention (DLP) solutions via the industry-standard ICAP. Send decrypted traffic to DLP servers for inspection before forwarding intercepted traffic to a client or a server.Gain Superior Visibility and Control into Application TrafficIdentify and categorize traffic on the application level, allowing for more granular controls and policies to be defined, with application visibility and control. This DPI-based service provides application visibility with comprehensive user and group awareness, providing deep insights into network traffic. Understanding application traffic trends in enterprise networks allows for effective security planning and sanctioning of allowed business applications.A10 Thunder CFW DeviceInternetClientSecure Web Gateway Protects the Enterprise PerimeterDeploy Thunder CFW, with integrated SSL Insight technology, to decrypt traffic for a variety of security products, including inline, non-inline (passive/TAP) and ICAP-enabled devices.clouds, offering unprecedented telemetry as well as 100 percent RESTful API coverage. The product supports multi-tenancy features like application delivery partitions (ADP) for segmentation.7650 CFWThunderby the Numbers370Gbps8 MLayer 4 CPS128KRules384 MConcurrent SessionsThunder CFWThunder CFW IPsecData Center-to-Data CenterVPNVPNBy unifying IPsec VPN, firewall and application delivery controller (ADC) capabilities,organizations are able to both load-balance traffic and protect the data center, services and related applications from DDoS attacks and other threats.Thunder CFW provides industry-leading high performance as a physical, virtual or containerized solution. The physical appliance with hardware acceleration supports scalable and high-performance on-premises deployments. For NFVi and private cloud deployments, the virtual appliance works with leading hypervisors such as VMware ESXi, KVM and Microsoft Hyper-V, and integrates with leading NFV-MANO solutions including Ericsson Cloud Manager, NEC Netcracker HOM, Cisco NSO and Red Hat OpenStack andmore. For flexible and efficient cloud native deployments such as Docker and Kubernetes, the container option can be used.All Thunder CFW options run on A10’s ACOS software, providing feature parity, regardless of form factor, which helps simplify and consolidate operations in any deployment environment.Flexible Deployment OptionsHardware specifications and performance numbers are subject to change without notice, and may vary depending on configuration and environmental conditions. As for network interface, it’s highly recommended to use A10 Networks qualified optics/transceivers to ensure network reliability and stability.*1 Tested in single appliance SSLi deployment with maximum SSL option. Cipher "TLS_RSA_WITH_AES_256_CBC_SHA" with RSA 2K keys are used for RSA cases, "TLS ECDHE_RSA_WITH_AES_128_CBC_SHA256" with EC P-256 and RSA 2K keys are used for PFS case. | *2 With maximum SSL | *3 With base model | *4 Optional RPS available | *5 10Gbps speed only | *6 Fixed SFP+ optical ports with dual rate (10GBASE-SR and 1000BASE-SX) | *7 Hardware Bypass model comes equipped with the hardwareTLS acceleration | ^ Certification in process | + FIPS model must be purchasedHardware specifications and performance numbers are subject to change without notice, and may vary depending on configuration and environmental conditions. As for network interface, it’s highly recommended to use A10 Networks qualified optics/transceivers to ensure network reliability and stability.*1 Tested in single appliance SSLi deployment with maximum SSL option. Cipher "TLS_RSA_WITH_AES_128_CBC_SHA" with RSA 2K keys are used.*2 With maximum SSL*3 With base model. Number varies by SSL model+ Thunder 14045 comes with a splitter cable for console to provide access to both modules. | ^ Certification in process* Carrier-class firewall along with CGNAT and data-center firewall with ADC are the common use cases. SSL Insight (SWG) and IPsec are not adequate use cases. *******************************************************************************************.TestedwithUDPtrafficforCGNATusecase.Visibility and Analytics with Harmony Controller• CGNAT-Subscriber session insights-Session opening and closing rates-TopN flow consuming subscribers-TopN bandwidth consuming subscribers-Subscriber user quota alerts-CGN resource tracking-Mapping distribution per protocol and per technology-NAT IP pool utilization-Session distribution per NAT technology• Firewall-Firewall rule performance and rule distribution by protocol-Top firewall rules by state-GTP traffic insights, roam-in/out view and policy violation analytics-Comprehensive transduction log with source/destination IP, port, protocol, application, application category, and firewall actions for better visibility and faster troubleshooting• Application-Application distribution by category-Top destination IP by application distribution-Bytes consumed by application category Application-aware Firewall• Recognition of thousands protocol signatures with identification of service types within applications• Support custom rules that run real-timeIP Threat Intelligence• Prevents malicious traffic from entering your network, based on customizable risk score and toleranceThreat Investigator• Rich and contextual analytics for object under investigation Operation Modes• Transparent forward proxy• Explicit forward proxy• Proxy chainingCentralized Management and Analytics with Harmony Controller• Device and configuration management across multiple sites• Deployment wizard with guided configuration• Centralized security policy management and enforcement• Rich TLS/SSL traffic and decryption analytics for traffic insight, application insight, URL insight, source and destination insight and more• Threat investigator view• Session log drill-down• Troubleshooting tools• Extensive logging for auditADC• Advanced L4/L7 server load-balancing-Full HTTP proxy, Fast HTTP, HTTP/2, FIX, SIP, RADIUS, FTP, TCP, UDP and more -High-performance, template-based L7 switching with header/URL/domain manipulation-Comprehensive L7 application persistence support• DNS load balancing-Layer 4 (TCP, UDP) and Layer 7 (DNS-UDP, DNS-TCP-DNS over HTTPS (DoH), DNS over TLS (DoT)-Recursive DNS-DNS firewall/RPZ-DNS cache• Global Server Load Balancing (GSLB)• Comprehensive IPv4/IPv6 support• aFleX® TCL-based scripting: deep packet inspection and transformation for customizable, application-aware switching• HTTP acceleration: HTTP connection multiplexing (TCP connection reuse),• RAM caching, HTTP compression• SSL acceleration: Hardware SSL, TLS 1.2, TLS 1.3 support, Elliptic Curve Diffie-Hellman Exchange (ECDHE) and other PFS ciphers• ACME client support for automating TLS certificate renewal with Let’s Encrypt, Sectigo and else • Dynamic threat intelligence feed updated in near real time• 30-plus public, private and proprietary sources to block “call homes” to command and control servers, identify known attack sources and mitigate zero-day attacks.Scalable, High-performance Platform• Advanced Core Operating System (ACOS)-Multi-core CPU support-Linear application scaling with scalable symmetrical multi-processing (SSMP) architecture-ACOS on data plane• Linux on control plane• IPv6 feature parity• Flexible traffic acceleration (FTA) for scalable flow distribution, common attack mitigation-Hardware FTA utilizing FPGAs• Scale-out clusterNetworking• Integrated Layer 2/Layer 3• Transparent mode/gateway mode• Virtual wire interface support• Routing — static routes, IS-IS (v4/v6), RIPv2/ng, OSPF v2/v3, BGP4+• L2 protocols (STP, RSTP, MSTP)• VLAN (802.1Q)• Link aggregation (802.1AX), LACP• Access control lists (ACLs)• Traditional IPv4 NAT/NAPT• IPv6 NAPT• Jumbo frame support*• Hardware-accelerated VXLAN*• NVGRELearn More About A10 Networks Contact Us/contact ©️2023 A10 Networks, Inc. All rights reserved. A10 Networks, the A10 Networks logo, ACOS, Thunder, Harmony and SSL Insight are trademarks or registered trademarks of A10 Networks, Inc. in the United States and other countries. All other trademarks are property of their respective owners. A10 Networks assumes no responsibility for any inaccuracies in this document. A10 Networks reserves the right to change, modify, transfer, or otherwise revise this publication without notice. For the full list of trademarks, visit: /a10trademarks.Part Number: A10-DS-15112-EN-23 Feb 2023• Multi-tenancy with application delivery partitions (ADP) with L3-7 separation• L3 virtualizationDevOps Tools and Integration• Ansible modules and playbooks• Terraform Thunder provider• HashiCorp’s Consul and Network Infrastructure Automation (NIA) integration• Thunder Kubernetes Connector (TKC)• OpenStack Octavia driver• Cloud-init support for auto-configuration on-OpenStack-OCI-AWS-Azure• Prometheus integration for visibility and analytics monitoring* Features and certifications may vary by appliance.** Additional paid service.。
E DGE I RONFoundry Networks’EdgeIron 2402CF,EdgeIron 4802CF, EdgeIron 24G,and EdgeIron 48G deliver a cost-effective solution for the enterprise wiring closet,small backbone,server farm,and desktop applications for high-performance local area networks.The EdgeIron family of 10/100 and 10/100/1000 switches delivers interface flexibility,wire-speed performance, superior port density,and a complete,standard Layer 2 feature set to address the mid-range needs of enterprise users.The EdgeIron products come in a 1-R U form factor,enabling seamless installation within a wiring closet,distribution,and data center.The EdgeIron products include combo Gigabit Ethernet ports that allow for RJ-45 (copper) or mini-GBIC (fiber) uplink installations.EdgeIron 2402CF,4802CF,24G,and 48G support mini-GBIC optics,such as SX for multimode fiber,LX for single-mode fiber,or LHA for single-mode fiber that can reach up to 70Km.L a y e r2S w i t c h e sFeatures/ Benefits Highlightst Cost-effective solution for high-performance LAN environments ina 1-RU form factort Wire-speed performance,auto-sensing,and auto MDI/MDIXon all portst Combo Gigabit Ethernet ports to deliver integrated RJ-45 (Copper)or mini-GBIC (fiber) for uplink installation,giving greater flexibility and cost savings for wiring closet installationst Support for jumbo frames of up to 9,000 bytes within the EdgeIron24G and 48G,ideal for high-end server connectivity and network attached file serverst Complete,standard layer 2 features that includes the following:—IEEE 802.1q and 802.1p (Class of Service) with 4 hardware queues per port to enable prioritization of mission-critical applications—Per-VLAN Spanning T ree (PVST) for broadcast isolation—802.3ad for automatic link aggregation and 802.1x for port security —802.1w Rapid Spanning T ree Protocol for superior network reliability —Support for Generic V LAN Registration Protocol (GVRP)—Internet Group Management Protocol (IGMP) Snoopingt Robust Quality-of-Service features,including Class of Service (802.1p) mapping to T ype of Service or DiffServ and support for priority queuing algorithm such as W eighted Round Robin and Strictt Low latency – as low as 10µs,ideal for advanced applications like VoIP and video conferencing over IPt Extensive management and monitoring features,including an industry-standard CLI,secure web-based GUI,integrated SNMP agent with mini-RMON and Secure Shell for secured and encrypted management accessEdgeIron 2402CFEdgeIron 4802CFEdgeIron 24GEdgeIron 48GThe EdgeIron 2402CF features 24 10/100 RJ-45 ports and 2 combo GbE RJ-45/mini-GBIC slots in a 1-R U form factor,ideal for wiring closet installation.The EdgeIron 2402CF comes with switching performance of up to 8.8Gbps and forwarding rate performance of up to 6.6Mpps.The EdgeIron 4802CF features 48 10/100 RJ-45 ports and 2 combo GbE RJ-45/mini-GBIC slots in a 1-R U form factor,ideal for higher-density wiring closet installation.The EdgeIron 4802CF comes with a switching performance of up to 13.6Gbps and forwarding rate performance of up to 10.2Mpps.The EdgeIron 24G features 20 10/100/1000 RJ-45 ports and 4 combo GbE RJ-45/mini-GBIC slots in a 1-R U form factor,ideal for wiring closet,distribution,and data center installation.The EdgeIron 24G supports jumbo frames,making it an ideal solution for server connectivity and network attached file servers.The EdgeIron 24G comes with switching performance of up to 48Gbps and forwarding rate performance of up to 35.7Mpps.The EdgeIron 48G features 44 10/100/1000 RJ-45 ports and 4 combo GbE RJ-45/mini-GBIC slots in a 1-R U form factor,ideal for wiring closet,distribution,and data center installation.The EdgeIron 48G supports jumbo frames,making it an ideal solution for high-density server connectivity and network attached file servers.The EdgeIron 48G comes with switching performance of 80Gbps and forwarding rate performance of up to 60Mpps.EdgeIron P roduct F amily K ey F eatures a nd B enefitsPerformance, Scalability and SecurityEnterprise networks require cost-effective performance, scalability,and high density to support mission-critical business applications and rapid network expansion.With wire-speed switching performance,255 active V LANs (each V LAN ID value may range up to 4,096),the EdgeIron family provides best-in-class performance and capacity.Enterprise network managers,who demands security,can rely on the “port-security”feature that comes with the EdgeIron 2402CF,4802CF,24G,and 48G.Enterprise network managers can restrict access to their network by allowing only secured hosts (using the MAC addresses) to a port.Additional security actions such as disabling a port can be activated when a “port-security”violation occurs,thus proactively eliminating security violations.Cost-Effective SolutionThe EdgeIron 2402CF,4802CF,24G,and 48G are ideal for providing cost-effective,high-performance connectivity in a variety of network deployments.The EdgeIron 4802CF is well suited to wiring closet deployments where high-density 10/100 is required.Key applications for the EdgeIron 24G and 48G include high-performance server aggregation,high-speed workgroup backbone upgrades,and Gigabit over Copper to the desktop for power users.Foundry Networks offers a complete line of complementary Layer 2/3 stackable and chassis products for the edge and the core of the enterprise,including the FastIron enterprise product family,that can be deployed in conjunction with the EdgeIron 2402CF,4802CF,24G,and 48G to create a complete,robust,scalable,and world-class high-performance network.Includes Complete Standard Layer 2 Feature SetEach EdgeIron ships with a complete,standards-based Layer 2 feature set,including 802.1Q VLANs,802.1p-based Class of Service (CoS),802.1D single Spanning T ree Protocol,Per V LAN Spanning Tree (PVST),802.1w Rapid Spanning Tree Protocol, 802.3x-based flow control,BootP,TFTP,Generic VLAN Registration Protocol (GVRP),Generic Multicast Registration Protocol (GMRP),Internet Group Management Protocol (IGMP) Snooping,and port mirroring.With support for CoS mapping to T ype of Service (T oS) and DiffServ Control Points (DSCP),the EdgeIron delivers critical Quality of Service (QoS) features for next-generation applications such as V oice over IP and streaming audio or video.Ease of ManagementWith an easy-to-use,industry-standard Command Line Interface (CLI),T elnet based interface,W eb based GUI,and RADIUS-based authentication,the EdgeIron is easy and secure to configure,deploy and maintain.Foundry’s full-function IronView Network Manager empowers network operators to seamlessly control software and configuration updates for any Foundry product from a central station,dramatically simplifying network provisioning,diagnostics and resolution,thus reducing the total cost of ownership of installing and maintaining a world-class network infrastructure.Building E nterprise N etworks U sing t he E dgeIronEnterprise Wiring Closet and DistributionThe EdgeIron 2402CF and 4802CF ,which come with 2 Gigabit Ethernet over Copper ports as uplinks,are well suited for enterprise wiring closet deployments and can be immediately connected to an EdgeIron 24G,which can be used within the distribution layer.With this solution and the EdgeIron’s 1R U size,network managers can easily collapse several EdgeIron 2402CF and 4802CF switches into an EdgeIron 24G and a 48G to delivera high-density wiring closet solution.Furthermore,the EdgeIron 24G and the 48G supports up to 4 mini-GBIC ports allows network managers to easily connect the distribution back to the enterprise backbone using SX,LX,or LHB optics,enabling distances of up to 300m,5Km,or 120Km,respectively .This solution delivers a cost-effective and efficient aggregation of wiring closets within an enterprise network.F astIron 800F astIron 800Enterprise BackboneWiring ClosetEdgeIron 2402CFEdgeIron 4802CFVLAN 100VLAN 200EdgeIron 24GDistributionLayerVLAN 100VLAN 200VLAN 100VLAN 200LegendGigabit LinksDual GbE Trunk LinksBuilding #1First FloorSecond FloorBuilding #2First FloorSecond FloorWiring ClosetEdgeIron 2402CFEdgeIron 4802CFVLAN 100VLAN 200EdgeIron 48GVLAN 100VLAN 200VLAN 100VLAN 200Enterprise Dat a CenterThe EdgeIron 24G and the 48G are well suited for use within the enterprise data center.The EdgeIron 24G and the 48G,which comes with 24 GoC and 48 GoC,respectively ,enables network managers to easily connect high-end servers with Gigabit Ethernet over Copper (GoC) ports.Both the EdgeIron 24G and the 48G comes with 4-port combo mini-GBIC slots,which can be populated with SX,LX,or an LHB optic to connect the data center back to the enterprise core at distances up to 300m,5Km,or 120Km,respectively .This solution delivers a cost-effective and flexible solution for data center networking requirements.Enterprise BackboneLegendDual GbE Trunk LinksData CenterServers with Gigabit Ethernet over CopperF astIron 800F astIron 800EdgeIron 24G EdgeIron 48GOSPF VRRPGigabit Copper to the DesktopT oday ,laptops and desktop computers are being equipped with 10/100/1000 Base-T Ethernet ports because the cost of adding this feature is negligible.Foundry’s EdgeIron 24G and 48G delivers a cost-effective solution for enterprise network managers wanting to future-proof their network to support 10/100/1000Base-T Ethernet ports.The EdgeIron 24G and the 48G comes with the ability to auto-sense 10/100/1000 Base-T ,eliminating the need to manually configure the speed of every switch port used for host connectivity .A wiring closet with several EdgeIron 24G and 48G connected to a FastIron 3208RGC,equipped with 32-ports of Gigabit Ethernet over Copper and 8 mini-GBIC slots,delivers a high-density and a cost-effective solution for future-proofing a wiring closet to support hosts with 10/100/1000 Base-TEthernet ports.With the FastIron 3208RGC and its additional 8 mini-GBIC slots,network managers can ensure redundant and high-speed connectivity of wiring closets to the enterprise backbone.F astIron 800F astIron 800Enterprise BackboneWiring ClosetVLAN 100VLAN 200Wiring ClosetEdgeIron 24GEdgeIron 48GVLAN 100VLAN 200DistributionLayerVLAN 100VLAN 200VLAN 100VLAN 200VLAN 100VLAN 200VLAN 100VLAN 200LegendDual GbE Trunk Links Quad GbE Trunk LinksBuilding #1First FloorSecond FloorBuilding #2First FloorSecond FloorEdgeIron 24GEdgeIron 48GTechnical SpecificationProperties EdgeIron EdgeIron EdgeIron EdgeIron2402CF4802CF24G48G10/100 Base-T Ports2448001000 Base-T Ports220010/100/1000 Base-T Ports002448Mini-GBIC Slots2244Switching Performance (Gbps)8.813.64880Forwarding Rate (Mpps) 6.610.235.760Maximum Number of MAC Addresses8,0008,00032,00016,000Physical Dimensions (HxWxD) 1.7" x 17.37" x 9" 1.7" x 17.37" x 12.76" 1.7" x 17.37" x 13.9" 1.7" x 17.37" x 13.9"(4.3 x 44 x 22.8 cm)(4.3 x 44 x 32.4 cm)(4.3 x 44 x 35.3 cm)(4.3 x 44 x 35.3 cm)W eight 7 lbs (3.18 Kg)9.5 lbs (4.36 Kg)9.5 lbs (4.36 Kg)9.5 lbs (4.36 Kg) Power Consumption (W att)24W Max48W Max48W Max48W MaxTechnical S pecification (Common)Network Interface-10BASE-T:RJ-45 UTP Cat.3,4,5 -100BASE-TX:RJ-45 UTP Cat.5 -1000BASE-T:RJ-45 UTP Cat.5 Unicast and Mul ticast Latency Performance (EIF2402CF,EIF4802CF, andEIF24G)-Under 10µs for Unicast port to port Latency-650 – 700µs Multicast Join-6.5 – 7s Multicast Leave Temperature-Operating:0 – 50˚C-Storage:-40 – 70˚CHumidity-10% to 90% (non-condensing)AC Input-100 to 240V,50-60 Hz Power Suppl y-Internal,auto-ranging transformer:90 to 260 V AC,47 to 63 Hz-Redundant DC inputMaximum Current-0.5A @ 110VAC,0.7 A @ 240VACMTBF-6 years (estimated)Switch Features-Spanning T ree Protocol and 802.1w-Flow Control (802.3x)-Full and Half Duplex;AutoMDI/MDIX-VLAN Support,Up to 256 V LANs;port-based or with 802.1Q V LANtagging-GVRP for automatic V LAN learning-Quality of Service (802.1p)-Supports four levels of priority andweighted fair queuing-Broadcast storm control-Link Aggregation (802.3ad)-Port MirroringManagement Features-In-Band Management-T elnet,W eb-based HTTPS,orSNMP (v1 and v2c)-Out-of-Band Management-RS-232 DB-9 console port-Software Loading-Secure Shell v 1.5MIB Support-MIB II (RFC 1213)-Bridging MIB (RFC 1493)-Ethernet-like MIB (RFC 1643)-SNMP (RFC 1157)-ARP (RFC826)-IEGMP (RFC1157)-IGMP (RFC1112)-Bridge MIB (RFC 1493),-RADIUS (RFC2618)RMON Support(RFC 1757)-Groups 1,2,3,9 (Statistics,History,Alarm,Event)Ordering InformationPart NumberDescriptionEIF2402CF 24 10/100 Base-T (RJ-45) ports and 2 combo Gigabit Ethernet RJ-45/mini-GBIC ports EIF4802CF 48 10/100 Base-T (RJ-45) ports and 2 combo Gigabit Ethernet RJ-45/mini-GBIC ports EIF24G 24 10/100/1000 Base-T (RJ-45) ports and 4 combo Gigabit Ethernet RJ-45/mini-GBIC portsEIF48G 48 10/100/1000 Base-T (RJ-45) ports and 4 combo Gigabit Ethernet RJ-45/mini-GBIC portsE1MG-SX 1000BaseSX mini-GBIC optic MMF ,LC connector E1MG-LX 1000BaseLX mini-GBIC optic SMF ,LC connector E1MG-LHB1000BaseLHA mini-GBIC optic SMF ,LC connectorStandards-IEEE 802.3 10BASE-T specification -IEEE 802.3u 100BASE-TX specification-IEEE 802.3ab 1000BASE-T specification -IEEE 802.3ad-IEEE 802.3z 1000BASE-X specification-1000BASE-X (SX,LX,and LHA mini GBIC optic for LC connectors) -1000BASE-T (RJ-45)-IEEE 802.1w Rapid Spanning T ree Protocol-IEEE 802.3x full duplex on 10BASE-T ,100BASE-T ,and 1000BASE-T ports-IEEE 802.1D Spanning T ree Protocol-IEEE 802.1p CoS prioritization -IEEE 802.1Q V LAN-IEEE 802.1x for Port SecurityISO and Compliances-IEC 8802.3-CE MarkSafety-CSA/NRTL (UL1950,CSA 22.2.950) -TUV/GS (EN60950)Electromagnetic Compatibility-CE mark-EN55022 (1997) Class A.-EN55024 (1998)-EN61000-4-2/3/4/5/6/11 -EN61000-2-2 Class A.-EN61000-2-3 -FCC Class A -VCCI Class A -CISPR Class A© 2004 Foundry Networks,Inc.All Rights Reserved.Foundry Networks,BigIron,FastIron,NetIron,ServerIron,IronPoint,Foundry Networks,Inc.Corporate Headquarters 2100 Gold Street P .O.Box 649100San Jose,CA 95164-9100U.S.and Canada T oll-free:1-888-TURBOLAN (887-2652)T el:+1 408.586.1700 Fax:+1 408.586.1900info @www Although Foundry has attempted to provide accurate information in these materials,Foundry assumes no legal responsibility for the accuracy or completeness of the information.More specific information is available on request from Foundry .Please note that Foundry’s product information does not constitute or contain any guarantee,warranty or legal binding representation,unless expressly identified as such in duly signed writing.。
摘要近年来,随着宽带网络的发展和多媒体技术的进步,以及用户对网上多媒体信息需求的增长,视频流媒体技术受到了越来越多的关注。
为了提高压缩效率和增加网络友好性,H.264视频编码标准引入了一些新特色,其中包括新编码类型SP帧(和sI帧),sP帧(SP,SwitchingPrcdictive—frame)的主要特色:不同参考帧可以解码出相同重构帧。
因此利用sP帧技术生成的码流,能够在不同码率的码流间进行无缝切换,从而服务器能够根据用户的连接速度提供最好质量的服务。
本文论述了一个基于sP帧技术的码流无缝切换系统的设计与实现。
在该系统中,首先,对SP帧的性质和编解码方寨做了深入研究。
假定同一个序列以不同的码率编码成两个码流.在每个码流中切换点位置处的图象编码成SP帧,称为PrimarySP帧,对于每个PrimarySP帧,有个相应的SecondarySP帧,这个SecondarySP帧与PrimarySP帧有等同的重构帧,SecondarySP只在码流切换时发送(称为切换帧)。
H.264标准的JM模型只实现了PrimarySP帧的编解码,~次编码仅完成一条码流,而对SecondarySP帧的编码环境设置和编码算法都没有实现。
本文从JvT提案描述的SecondarySP帧的粗略框架出发,围绕“无缝切换”的中心目的,对图像的帧内宏块以“拷贝”的特殊形式编码.帧间宏块以新框架编码,同时对特殊的SKIP编码模式考虑适用性,以及滤波器的更新等,详细设计了一个SecondarySP帧的编码算法。
第二,针对“码流切换”的研究重点,本文提出了一种高效的智能流文件框架。
从“快速,准确无误”的根本出发,对不同码率的基本码流块,码流间的切换帧块,关键帧块,在文件中进行合理安排。
此框架所产生的智能流文件结构清晰,管理方便,适用于常见流媒体服务器。
结台上述两个技术,最终实现了“~次编码可完成多条码流”且“多条码流可无缝切换”的系统。
Fast Scalable File Distribution Over Infiniband*Dennis Dalessandro Ohio Supercomputer CenterSuite 3101 S. Limestone StSpringfield, OH 45502dennis@Pete WyckoffOhio Supercomputer Center 1124 Kinnear RoadColumbus, OH 43212pw@AbstractOne of the first steps in starting a program on a cluster is to get the executable,which generally resides on some network file server.This creates not only contention on the network,but causes unnecessary strain on the network file system as well,which is busy serving other requests at the same time.This approach is certainly not scalable as clusters grow larger.We present a new approach that uses a high speed interconnect,novel network features,and a scalable design.We provide a fast,efficient,and scalable solution to the distribution of executable files on production parallel machines. KeywordsRDMA, Scalable File Distribution, Infiniband, File system, Cluster1. IntroductionOften times in the cluster environment of today a program is executed on a cluster by a number of nodes receiving the executable code from a single source.Sun's Network File System(NFS) [14]is a prime candidate for such a task as it is popular, stable, and widely available. As executables grow larger, and the number of nodes increases,the single NFS server struggles to handle the additional load, as it was not designed to handle these types of situations. Not only must NFS send more data to more nodes, it also has to be available to service file requests from other users on the cluster.Two immediate observations show why using NFS is the wrong approach for executable distribution. First, the parallel processes are cooperating and all need exactly the same set of files at exactly the same time. Having each node request the file from NFS independently causes unnecessary load on the server and much duplicated effort among the clients.Second,high performance clusters generally have some sort of fast interconnect for message passing traffic,such as Myrinet[21],Quadrics[16],or Infiniband[1].However, installations generally use a fast or gigabit Ethernet network for file system traffic, perhaps because using NFS generally requires using an IP stack that may not be supported on the high performance network or on the file system server.Our approach to these problems starts by recognizing the collective nature of parallel executable startup.Collective operations appear in many contexts. In message passing, the MPI library [5] offers primitives for functions such as broadcast and reduce that optimize the data transfers when compared to the equivalent collection of point-to-point primitives.In peer-to-peer file *Support for this project was provided to the Ohio Supercomputer Centerthrough the Department of Energy ASC program .1sharing, collectives are used to save link bandwidth at the source host and fully utilize all network paths among downloading hosts.The parallel startup problem is similar to both of these,but the nature of the interconnection network and the types and sizes of files lead us to a particular solution.Modern networks offer features that can be used to enhance message passing performance.We investigate the advantages and drawbacks of using Remote Direct Memory Access(RDMA) [2]to perform the file distribution.It allows us to move data between nodes without the involvement of the operating system or host processor, exposing opportunities for overlap during the startup phase.In Section 2, we have a brief introduction to Infiniband [1] and RDMA [2], followed by a discussion in section3of the environment that Fast Startup is intended to take part.In section4we focus on the basic layout and the decisions which influenced the engineering of the Fast Startup software. Next in section 5 we turn to experimental results to show that Fast Startup is indeeda better choice. Section 6 looks at related work and we conclude in section 7.2 Infiniband and RDMAInfiniband is a technology to interconnect processor and I/O nodes to form a system area network [1]. It is a switched, point-to-point network with high speed data transfer and low latency. RDMA [2] is a technology that allows the transfer of data directly from a process running on a node into the memory of a process running on a remote node. The OS and CPU are not involved in this transfer.This is what makes RDMA so appealing.Infiniband provides both RDMA read and write data transfers.As one of our target clusters uses 10 gigabit Infiniband, we choose that as the initial implementation platform. Other alternatives include Quadrics [16] and RDMA over Ethernet(iWarp)[2],but those technologies are not yet as inexpensive or prevalent. However, as other interconnects mature, we want to take advantage of them. Thus, we were careful in designing the code to allow for this, as described in Section 4.3. Resource Management EnvironmentThe collective file distribution problem is constrained and orchestrated by the resource management environment. Production machines use software such as PBS [4], Platform LSF [17],or Slurm[18]to control the assignment of individual hardware elements in a parallel machine to particular ers submit job scripts that specify requirement s, such as number of processors and amount of memory, and a list of shell commands to run to perform the tasks. MPI jobs, in particular, use a job launcher such as mpirun [7] or mpiexec [6] to spawn the executable on each processor and initialize communications among the nodes. There are three phases to starting a parallel executable: signaling a daemon on each node to begin a task,launching the executables themselves, and initializing communications in the context of the new parallel job. The final phase, initializing communications among nodes, has been the subject of much previous work and is discussed briefly in Section 6.For the first phase, the MPI job launcher program contacts each node in turn to instruct it to fork and execute one task of the parallel process. Using2MPICH's [19] mpirun [7], this involves using rsh or ssh in a loop to contact the inetd on each node to start the tasks. With LAM [15], MPD [10], and mpiexec [6], an existing collection of daemons are used to start up the processes efficiently. In the case of mpiexec,these daemons are the individual resource managers themselves. Regardless, this signaling process is relatively lightweight in terms of resource consum ption on the network and scales well.The second phase generally has each process independently invoke the exec() system call directly on a shared file system, such as NFS [14], to start the executable.As discussed earlier,this activity is a major limitation to scalability.Besides the executable itself,shared libraries associated with the code also must be read into the memory of each node at this time. The job may further specify common input files to be used by all the processes. Distributing all these files collectively on the message passing network is the subject of the next section.4. DesignThe basic design of Fast Startup consists of three components.First, there is a transport specific module that currently handles all the transferring of data from a high level view,such as sending and receiving a file.There is another module which handles helper functions for the transport module, such as initializing the Infiniband stly,we have the main module that oversees all operation of the application.It handles making connections, coordinating sending and receiving, and everything else related to the parallel distribution of the file.4.1 APIThe API we have chosen to use is VAPI [8] which is the lowest level API that enables us to interface directly with the Infiniband hardware.This is a verbs API, and is the least common denominator on Infiniband systems.We have chosen not to use something like DAPL [3] because it is not very common yet. Other API's such as IPoIB [20] do not expose the necessary function calls to do RDMA directly.4.2 TopologyOne of the most important design considerations is to use a tree-based approach for the distribution of the files.This way we have aΟ(log N) algorithm. Taking a step back we can see this is already a better approach and the efficiency of the distribution will out perform the traditional NFS [14]file distribution, which happens to scale linearly as we show in section 5.Things can only get better by increasing the performance of the network. One of the aspects we wanted to look at is the effect of arity, or the maximum number of nodes one node in the tree may send to. It was important to make this an option that can be set at runtime.For extremely large numbers of nodes, it may be better to allow nodes to send to more than just two others, so instead of a binary tree requirement we have an n-ary tree flexibility.The biggest advantage of the tree based distribution is that it is easily scalable, which is an absolute requirement for this research.34.3 Communication ManagerIn the future,porting this software to DAPL[3]to make use of a communication manager(CM) could be quite useful.This would allow for a reduction in the connection build up time between nodes that we have to endure at the present time. More on this follows. As of right now there is no accepted standard for a communication manager on Infiniband so we should not rely on any one in particular.Since we are not using DAPL [3] we use TCP as a means of building up the initial connection between two nodes.This amounts to sending Infiniband connection related information which can be handled more efficiently when using a communication manager.We show the portion of time it takes to complete this later.4.4 File TransferOne of the major advantages offered by this work is the utilization of advanced networking features,namely Remote Direct Memory Access(RDMA). There are, however, limitations on the use of this mechanism, and thus we have identified three different schemes to send a file. Those are mmap all, multi send multi map,and multi send one map.All three use RDMA to move data around. The differences are in how the memory is allocated and if the file can be sent at one time or not.Presently, Infiniband allows a maximum transfer of 2GB at one time [9]. Clearly we need a way to send larger files since this is meant to be scalable in every aspect.It may also be desired that a limit be set on the maximum size that can be sent at one time to take advantage of pipelining.One other important configuration choice is the chunk size.Chunk size refers to the amount of data that can be sent at one time. All of these factors significantly impact performance as we will see later.Mmap all, as the name implies, involves memory mapping the entire file. On the Root Node the file is opened through the regular OS method and mmap() is used to map the file to the send buffer. There are no memcpy()calls made. On the receiving node an empty file is created and mapped to its receive buffer. The RDMA that occurs moves the file directly from the local disk of the root node to the local disk of the remote node. Now the node that just received the file can open the new local file and map it in.Multi-send multi-map, unlike the above,maps only portions of the file and copies the data to a buffer to send.This type of send involves multiple memory maps to be made. This is essential to support large files. It may be the case that the file being transferred is too large to fit into a contiguous area of memory, thus requiring multiple memory maps.Similarly,multi-send one-map sends the file in chunks. The difference is that multi-send one-map will map the entire file on both ends like with mm ap all. This achieves the greatest performance, but the available memory may not always be available to hold the entire file at one time.In both of the multi send methods the sender will poll its completion queue[9]to verify the send request was properly formatted and wait for an acknowledgment from the remote node that it has received and processed the data.The remote node knows when data has been received by polling its4completion queue. It is an RDMA write with immediate data [9] that makes this possible. We show the performance implications of the three send types in the next section.We support all three send types in the software to ensure that Fast Startup can handle any size file.5. ResultsThe results that follow were gathered on OSC's P4 Cluster which includes 112 compute nodes connected with 10 gigabit Infiniband and 1 gigabit Ethernet. Each with two 2.4 GHz Intel P4 Xeon processors and 4GB of RAM. The file sizes used may seem quite large compared with the common executable.This, however, is intentional. It is meant to show the scalability of Fast Startup, and is motivated by the need to send shared libraries and input files. It should be noted that all data was gathered on a 2-ary (binary) tree, with the exception of the last data set which looks at multiple arities.5.1 Total Time to DistributeOne of the most important measures we need to look at is the overall time to distribute the file to a number of nodes.Having measured the mean time it takes to send a 100MB file, 361 milliseconds, and the mean time it takes to receive a 100MB file, 193 milliseconds (see table 1), we can know how long it will take to distribute amongst a varying number of ing the send history of runs of various size nodes, from 2 to 100 in increments of 3, we can plug in the mean send and receive times to calculate the total expected distribution time. We do this in software to account for the parallelism of a tree based distribution. We now have the most general case possible of sending to a certain number of nodes. The results are shown in Fig 1.The reason we do not measure this directly is because it is difficult to get a very accurate result without the notion of a global clock, which is not possible to achieve [22]. The root node is the only node that knows when the distribution started, and the only way for it to know when the distribution is completed is to have every node report when finished.This adds significant overhead and skews the result. The same situation arises in section 5.5.Fig 1. Time to distribute 100MB.5We see some flat areas on the graph, and that illustrates perfectly the advantage gained from a tree based algorithm. The flat areas are due to different numbers of nodes receiving the file in the same amount of sends. Ultimately the cause of this is the ability to have many sends and receives happening in parallel with each other.It is also good measure to compare this with how long it takes NFS to distribute to a certain number of nodes. Data in Fig 2 was generated on OSC's large mass storage system. This is comprised of multiple dual P4 2.4GHz with SCSI RAID disks and a 1Gbps Ethernet interface. These NFS servers fan out via high performance switches to connect to the compute nodes on the cluster. Due to the fact that NFS is so slow and inefficient we have only tested NFS up to 30 nodes. See Fig 2.We can see from Fig. 2 that it takes roughly 1.5-2 seconds per node. While with Fast Startup and Infiniband it takes roughly3seconds to distribute to100 nodes.Potential reasons for the poor performance of NFS include,lack of server software stack scalability and heavy load on the network.We see30 nodes take roughly70seconds to receive100MB each.This is an aggregate bandwidth of 43MB/s, well under the potential 125MB/s available in the gigabit Ethernet network.Therefore,we suspect the biggest drain on performance comes from the overheads associated with the NFS protocol and costly send/r eceive network semantics rather than problems with load on the network.5.2 Basic Send PerformanceWe next look at how long Fast Startup takes to send a certain amount of data. Using files ranging from 4KB all the way up to 1GB we show the rate at which Fast Startup can send the file to the remote node, including all overhead incurred by our software.Then the remote node saves the file locally and acknowledges the sender. The data in Fig. 3 represents the mean time to send the various file sizes. The mean was taken over a large number of trials. The standard deviations of these means were always less than 1%. We see that the mean is pretty accurate,and that the trend of the graph is very linear as we6would expect. The slope is linear across the entire range.A small 4KB executable took almost no time (0.3 milliseconds), while even the 1GB file takes less than 2 seconds. The mean time to send a 100MB file is 169 milliseconds, which we will look at as a general case through out.5.3 Chunk SizeAnother interesting performance aspect that we wanted to look at is the effect of chunk size. We can see what impact chunk size and the type of send has on the overall performance by looking at Fig. 4. The graph in Fig. 4 depicts the mean time to send a 100MB file. As with the above, this time includes the time to send the file to the remote node, then for the remote node to save the file locally and acknowledge the sender.From the above graph we see that a mmap_all is far superior to the other two. This is due to the overhead involved in each RDMA operation. The reason that7the mmap_all line is flat is that chunk size has no effect since the file is sent all at once with this type of send. The reason for even having the other two types of sends are for files larger than can be sent in one send through the interconnect.The multi-send multi-map type performs slightly worse than multi-send one-map.As we can see the two lines start to get closer and closer together. This is as expected because as we increase the size that we are mapping it is getting closer to one large mapping.Clearly we want to map as much as possible at one time, and minimize the number of mappings that we need to do in order to gain the best performance.The lesson to be learned here is that if a file can not be sent in one send due to restrictions,either on the interconnect or the restrictions on mapping memory,it is best to make the chunk size as large as possible and use the multi-send one-map type of send. Unfortunately mapping memory turns out to be the biggest problem due to the fact that it needs to map contiguous area of memory, so for large files, lots of memory is needed.5.4 Time BreakdownFor the mm ap_all case we break down the entire process of sending and receiving a file to see just where the time is spent and how much time is spent in each stage.Over many nodes the average to send a100MB file is361.3 milliseconds. This time is different than that shown in Fig. 3 as it is the total send time which includes all overheads.The components that make up this time are illustrated below in Table 1.As mentioned before, the time to build the connection is by far the most costly overhead. The time spent on the RDMA transfer, which includes setting up the transfer on the NIC and polling for the results, is the only useful work. The time to build the initial connection includes significant overhead due to the TCP stack, and rivals the time to actually send 100MB of data over Infiniband.There is a slight delay while the sender waits for the receiver's QP to be ready. This involves the receiver sending a message to the sender, as we see it8takes3milliseconds for this to happen.Similarly it takes another 3.4 milliseconds for the sender to send the initial RDMA which includes a list of nodes that still need the file,this includes waiting for the receiver to acknowledge it has received the initial RDMA. Another 5 milliseconds is spent maintaining data structures and tearing down the connections.Looking at the Receive section of Table 1, we see there is very little time wasted.Clearly receiving the file is more efficient because the TCP socket to transfer the initial Infiniband connection data is already created, and waiting on the sender. This is quite desirable because all nodes must spend time doing the receive,but not all must participate in the sending stage.The overheads involved in receiving the file are small(4.9and 5.1milliseconds),and are basically independent of file size.So most of the time is spent doing actual work to receive the file, actually it is passively waiting for data to be written into its buffer, rather than actively receiving the file.We hope to greatly speed up the connection management using a better interface such as DAPL [3].5.5 ArityThe remaining aspect we need to look at is arity.Arity as used here means the number of nodes a sender is able to distribute the file to.Fig.5 shows the effect of arity on distributing a100MB file to100nodes using the mmap_all type of send.From the graph we can see that a binary tree is actually the best performing type of distribution. While this may seem surprising at first, after looking at the total send and receive times, we see that the total send time per node is in fact quite a bit more than the total time it takes to receive per node. What we can tell from this is that we want to minimize the number of sends that occur sequentially and maximize the number of sends occurring in parallel.6. Related WorkAccording to[11]there are two phases to start a parallel application,9process initiation and connection set up.We propose there is in fact three phases,before a process can be started the executable has to reside locally. Other papers have been presented which focus on the two phases described in [11]. This is where Fast Startup comes in, and is what makes it different. MPD [10],is meant to be a fast way of starting parallel jobs once the file resides locally,and a flexible run time environment.In[11],the concession is made that to eliminate the impact of network communication to file access,all files were duplicated on local disks,further[11]mentions the file system performance could be a big bottleneck for a larger cluster. Efficiently starting the process once the file is there, is a vital issue to be addressed and therefore our Fast Startup work is by no means meant to be a replacement for earlier works, but in fact complimentary.Two novel systems that do take into account the need to get the file locally are yod[12]and STORM[13].They both address the two phases that [10] and [11] focus on, in addition they focus on what we propose as the third phase.While Yod[12] does handle all three phases it is not completely scalable, as it uses an O(N) file distribution technique. STORM [13], on the other hand, is scalable, and handles all three phases. The drawback with STORM [13] is that it requires the use of a hardware multicast mechanism specific to Quadrics[16] while we use only the more general RDMA operations.7. Conclusions & Future WorkWe have presented an efficient scalable method to distribute files among a number of cluster nodes in the context of a parallel process. We have shown there are three potential ways to manipulate buffers during file distribution, and we have looked at the effects of arity and chunk size on the total time to distribute the file. We have also shown that existing methods perform and scale poorly compared to our approach.Future plans for Fast Startup involve more extensive scaling studies on large cluster environments and using multiple network types.Potential improvements specific to Fast Startup include a study on using multiple trees to distribute the file among the nodes.Involving the specifics of the underlying network topology in the construction of the distribution trees may be necessary for extreme scale heterogeneous parallel machines.10References1. About Infiniband Trade Association: An InfinibandTechnology Overview./ibta/.2. RDMA Consortium website http://www.rd macons .3. DAT Collaborative website /.4. Open PBS /.5. MPI The message passing interface /m pi/.6. P. Wyckoff. Mpiexec. /~p w/m p iexec/.7. Mpirun /m pi/www/www1/m pir u n.html.8. Mellanox IB-Verbs API (VAPI) Mellanox Software Program m ers Interface forInfiniband Verbs, 2001.9. Infiniband Architecture Specification vol. 1 rel. 1.1, 6 Nov 2002.10. R. Butler, W. Gropp, and E. Lusk. Components and Interfaces of a ProcessManagement System for Parallel Programs.Parallel Computing, 27(11):1417 1429, 2001.11.W. Yu, J. Wu, D.K. Panda. Scalable Startup of Parallel Programs overInfiniband. Technical Report OSU-CISRC-5/04-TR33, Dept. of Computer and Information Science, The Ohio State University, Columbus, OH 43210, 2004.12. R. Brightwell, L.A. Fisk. Scalable Parallel Application Launch on Cplant.Proceedings of SC2001, Denver, Colorado, November 10 16, 2001. Available from /pa per s/p a p.pap263.pdf.13.E. Frachtenberg, F. Petrini, J. Fernandez, S. Pakin, S. Coll. STORM: Lightning-Fast Resource Management. In Proceedingsof the IEEE/ACM SC2002Conference, 2002.14. B. Callaghan, B. Pawlowski, P. Staubach. NFS Version 3 ProtocolSpecification. June 1995 available via /cgi-bin/rfc/rfc1813.ht ml.M/MPI Parallel Computing /.16. Quadrics corporate websitehttp://d /q u a drics/Q uad ricsHome.nsf/DisplayPages/Homepa ge.1117. PLATFORM LSF Intelligent, policy-driven batch application workloadprocessing — for desktops, servers and mainframes./p r o d ucts/LSF/.18.SLURM: A Highly Scalable Resour ce Manager/linux/slur m/.19.MPICH A Portable Implementation of MPI http://www-/m pi/m pich/.20. H.K. Jerry Chu, V. Kashyap. Transmission of IP over InfiniBand. August2004 available via /internet-drafts/d raft-ietf-ipoib-ip-over-infiniband-07.txt.21. Myrinet Home Page /.22. M. Singhal, N Shivaratri,Advanced Concepts in Operating Systems, Mc GrawHill 1994. p.98.12。