6 Storage Area Networks

目前磁盘存储市场上，存储分类（如下表一）根据服务器类型分为：封闭系统的存储和开放系统的存储，封闭系统主要指大型机，AS400等服务器，开放系统指基于包括Windows、UNIX、Linux等操作系统的服务器；开放系统的存储分为：内置存储和外挂存储；开放系统的外挂存储根据连接的方式分为：直连式存储（Direct-Attached Storage，简称DAS）和网络化存储（Fabric-Attached Storage，简称FAS）；开放系统的网络化存储根据传输协议又分为：网络接入存储（Network-Attached Storage，简称NAS）和存储区域网络（Storage Area Network，简称SAN）。

由于目前绝大部分用户采用的是开放系统，其外挂存储占有目前磁盘存储市场的70%以上，因此本文主要针对开放系统的外挂存储进行论述说明。

第一个图有问题，把NAS和SAN一样放在FAS之下是不对的，通常也没有FAS 这种说法，DAS，NAS和SAN是平行的关系。

NAS不一定要用光纤。

NAS是文件级存储，SAN和DAS通常是数据块级存储。

表一：今天的存储解决方案主要为：直连式存储（DAS）、存储区域网络（SAN）、网络接入存储（NAS）。

如下表二：开放系统的直连式存储（Direct-Attached Storage，简称DAS）已经有近四十年的使用历史，随着用户数据的不断增长，尤其是数百GB以上时，其在备份、恢复、扩展、灾备等方面的问题变得日益困扰系统管理员。

主要问题和不足为：直连式存储依赖服务器主机操作系统进行数据的IO读写和存储维护管理，数据备份和恢复要求占用服务器主机资源（包括CPU、系统IO等），数据流需要回流主机再到服务器连接着的磁带机（库），数据备份通常占用服务器主机资源20-30%，因此许多企业用户的日常数据备份常常在深夜或业务系统不繁忙时进行，以免影响正常业务系统的运行。

直连式存储的数据量越大，备份和恢复的时间就越长，对服务器硬件的依赖性和影响就越大。

SAN是Storage Area Network的缩写，也就是说SAN是一个网络；NAS是Network Attached Storage的缩写，也就是说NAS是一个存储设备；因此SAN和NAS根本不是同一类东西，因此根本无法将它们进行比较。

为什么有许多人要比较SAN和NAS，有以下两种情况：一，他们比较的是服务器是连接到Fabric网络（SAN）还是通过IP（LAN）网络连接到存储设备，前一种连接的是光线阵列，后一种连接的NAS设备。

为了卖出设备，因此一定要说出那个好，那个不好。

二，说明SAN和NAS互为补充，例如NAS后面使用SAN的网络作为存储。

比较fabric网络和以太网络：观点一：fabric网络性能高于以太网络，原因如下：1，从设计上，fabric网络就设计为高速传输的网络，2，在Fabric网络中，没有竞争，不需要确认，数据传输效率高，一个镇最大可谓2112字节。

3，Fabric是一个智能网络，自动重新路由，trunking（多端口绑定，带宽可达8Gb）4，Fabric网络中数据传输为块操作，因此对要求直接对磁盘能够读写的数据库有很好的支持能力。

而NAS一般不支持数据库。

观点二：FC网络性能再好，也是一个存储网络，服务器连接存储的性能再高，也是要对外提供服务的，服务器需要通过网络对外提供服务，你后端再快，也要受到前端网络的限制。

FC网络相当于使用了一个第二网络传输存储数据，如果使用第二个IP网络访问NAS 设备，因为第二网中机器少，因此效率肯定会大大提高。

FC网络目前带宽为1Gb，2Gb，即使通过trunk也只能达到8Gb，而万维网10Gb已开发出来并投入使用。

FC网络号称传输效率高，而IP网络如果使用cisco的0干扰交换机，高性能的千兆以太网卡，传输性能也很高。

FC的先天性缺陷就是数据共享能力差，如果要多台机器共享一个数据卷，需要sanergy，cvfs等软件的支持，而且管理信息也需要通过IP网络进行传输，而且不支持迁移等操作，性能也很差。

Unit Eight: The InternetUnit Eight/Section AI.Fill in the blanks with the information given in the text:1.research2.ICANN或the Internet Corporation for Assigned Names and Numbers3.router; gateway4.temporary/dial-up; permanent/dedicated5.ISP或Internet service providerwork; host7.decimal8.mnemonicII.Translate the following terms or phrases from English into Chinese and vice versa:1.cell phone 蜂窝电话，移动电话，手机2.IP address 网际协议地址，IP地址3.autonomous system 自主系统4.dial-up connection 拨号连接work identifier 网络标识符6.binary notation 二进制记数法7.mnemonic name 助记名，缩写名8.Internet-wide directory system 因特网范围的目录系统 server 名称服务器10.Internet infrastructure 因特网基础结构11.助记地址mnemonic address12.网吧cyber cafe13.宽带因特网访问broadband Internet access14.顶级域名top-level domain (TLD)15.因特网编址Internet addressing16.点分十进制记数法dotted decimal notation17.因特网服务提供商Internet service provider (ISP)18.专用因特网连接dedicated Internet connection19.主机地址host address20.硬件与软件支持hardware and software supportIII.Fill in each of the blanks with one of the words given in the following list, making changes if necessary:Early computer networks used leased telephone company lines for their connections.Telephone company systems of that time established a single connection between sender and receiver for each telephone call, and that connection carried all data along a single path. Whena company wanted to connect computers it owned at two different locations, the companyplaced a telephone call to establish the connection, and then connected one computer to each end of that single connection.The U.S. Defense Department was concerned about the inherent risk of this single-channel method for connecting computers, and its researchers developed a different method of sending information through multiple channels. In this method, files and messages are broken into packets that are labeled electronically with codes for their origins, sequences, and destinations. In 1969, Defense Department researchers in the Advanced Research Projects Agency (ARPA) used this network model to connect four computers into a network called the ARPANET. The ARPANET was the earliest of the networks that eventually combined to become what we now call the Internet. Throughout the 1970s and 1980s, many researchers in the academic community connected to the ARPANET and contributed to the technological developments that increased its speed and efficiency.IV.Translate the following passage from English into Chinese:因特网只是提供了将许许多多的计算机连接在一起的物理与逻辑基础结构。

存储基础知识DAS、SAN、NAS详解说明目前磁盘存储市场上，存储分类（如下表一）根据服务器类型分为：封闭系统的存储和开放系统的存储，封闭系统主要指大型机，AS400等服务器，开放系统指基于包括Windows、UNIX、Linux等操作系统的服务器；开放系统的存储分为：内置存储和外挂存储；开放系统的外挂存储根据连接的方式分为：直连式存储（Direct-Attached Storage，简称DAS）和网络化存储（F abric-Attached Storage，简称FAS）；开放系统的网络化存储根据传输协议又分为：网络接入存储（Network-Attached Storage，简称NAS）和存储区域网络（Storage Area Netw ork，简称SAN）。

由于目前绝大部分用户采用的是开放系统，其外挂存储占有目前磁盘存储市场的70%以上，因此本文主要针对开放系统的外挂存储进行论述说明。

表一：存储入门：图文阐释DAS、NAS、SAN（图一）今天的存储解决方案主要为：直连式存储（DAS）、存储区域网络（SAN）、网络接入存储（NAS）。

如下表二：存储入门：图文阐释DAS、NAS、SAN（图二）开放系统的直连式存储（Direct-Attached Storage，简称DAS）已经有近四十年的使用历史，随着用户数据的不断增长，尤其是数百GB以上时，其在备份、恢复、扩展、灾备等方面的问题变得日益困扰系统管理员。

主要问题和不足为：直连式存储依赖服务器主机操作系统进行数据的IO读写和存储维护管理，数据备份和恢复要求占用服务器主机资源（包括CPU、系统IO等），数据流需要回流主机再到服务器连接着的磁带机（库），数据备份通常占用服务器主机资源20-30%，因此许多企业用户的日常数据备份常常在深夜或业务系统不繁忙时进行，以免影响正常业务系统的运行。

直连式存储的数据量越大，备份和恢复的时间就越长，对服务器硬件的依赖性和影响就越大。

1.1直接连接存储服务器需要连接硬盘存储的一般的技术连接方式是直接存储(DAS)。

这种方式是给每一个服务器分配对应的硬盘存储。

虽然DAS的设置及购买实施很简单直接，我们很快意识到每一个新的存储投资只能对应于一个特定的服务器。

对于DAS存储的有效利用率及管理很快成为管理员的主要挑战。

图2直接连接存储(DAS )不同应用的服务器与存储一一对应在方便地设置DAS服务器的同时也就隐含了一些弊端。

在图1中我们可以看到不同服务器所对应相连的硬盘利用率是不同的。

有些服务器的存储利用率高达90%，而有的只有30%或40%。

如果将利用率乘以10或100，对于一些大的企业而言，他们的IT部门并没有对服务器的存储进行有效的利用。

进一步研究利用率的问题我们会发现，造成这个问题的原因是DAS的连接方式，不能使服务器之间共享存储资源。

DAS方案还有一个问题是服务器没有一个存储的中央管理，比如数据的备份及恢复，需要对每一个服务器独立进行。

服务器的存储容量的低利用率问题，某些服务器的存储空间用完，需单一管理一定数量的不能共享的存储资源，缺乏灵活性，所有的这些弱点导致浪费资源，潜在导致更高的管理压力及高的成本。

这些不足推动存储业急需开发一种新的存储技术去解决这些问题。

1.2网络化存储为了解决DAS的弱点问题，一种新的存储解决方案应运而生，将单一的存储整合为一共享存储池。

此方案可以给用户带来两大好处:降低成本及简化管理。

存储池的目的是在物理上将孤立的存储小岛整合成大片陆地，由此来提高存储的利用率。

也由此来减少不同的管理工具，有效地实现各种如数据备份的任务。

所有的这些好处将有效地降低客户的总成本。

现有的突出的存储整合方式有两种:Networked Attached Storage (NAS)和Storage Area Networks (SANs)。

虽然这两种方式都是实现了存储共享的目的，不过它们各自有自己的优点与弱点---取决于不同的应用与现有的IT环境。

什么是存储虚拟化什么是存储虚拟化那么什么是存储虚拟化呢？不同的公司和企业有不同的定义。

虽然虚拟化并不是⼀个全新的概念，但是在被引⼊到存储领域后却发⽣了某些变化，被赋予了新的内涵。

存储虚拟化是通过存储虚拟化的技术⽅法，将系统中各种异构的存储设备映射为⼀个单⼀的存储资源，对⽤户完全透明，达到互操作性的⽬的。

通过虚拟化技术，⽤户可以利⽤已有的硬件资源，把SAN内部的各种异构的存储资源统⼀成对⽤户来说是单⼀视图的存储资源（Storage Pool），⽽且采⽤Striping、LUN Masking、Zoning等技术，⽤户可以根据⾃⼰的需求对这个⼤的存储池进⾏⽅便的分割、分配，保护了⽤户的已有投资，减少了总体拥有成本（TCO）。

另外也可以根据业务的需要，实现存储池对服务器的动态⽽透明的增长与缩减，更进⼀步，可以实现SAN与SAN之间的虚拟化、全球的虚拟化。

虚拟化存储的能量正在释放存储技术经历了从单个的磁盘、磁带、RAID到存储⽹络系统的发展历程。

传统的直接存储(DAS)⽅式是存储设备附属于某个服务器，数据被局限在某个主机的控制之下，这种⽅式已远远不能满⾜企业分布式业务的需要，因⽽发展出⽹络存储技术。

典型的⽹络存储技术有⽹络附加存储(NAS，NetworkAttached Storage)和存储区域⽹(SAN，Storage Area Networks)两种。

NAS技术是⽹络技术在存储领域的延伸和发展。

它直接将存储设备挂在⽹上，具有良好的共享性、开放性;但缺点是与LAN共⽤同⼀物理⽹络，易形成拥塞⽽影响性能，特别在数据备份时性能较低，影响了它在企业级存储应⽤中的地位。

SAN技术的存储设备是⽤专⽤⽹络相连的，⽬前这个⽹络是基于光纤通道协议。

由于光纤通道的存储⽹和LAN分开，性能得到很⼤提⾼。

在SAN中，系统扩展、数据迁移、数据本地备份、远程容灾数据备份和数据管理等都⽐较⽅便，整个SAN成为⼀个统⼀管理的存储池（Storage Pool）。

PVC（Permanent Virtual Circuit）永久虚电路SVC（Switched Virtual Circuit）交换虚拟电路SNMP(simple network management protocol) 简单网络管理协议SGMP(simple Gateway Monitoring Protocol) 简单网关监控协议CA TV 公用天线电视和电缆电视PSE 分组交换机TDM((Time Division Multiplexing) 时分多路复用FDM(Frequency Division Multiplexing) 频分多路复用WDM(Wave Division Multiplexing) 波分多路复用PDSN 分组数据服务接点Console 控制台CGI(common Gateway Interface) 通用网关接口FDDI(Fiber Distributed Data Interface)光纤分布式数据接口HDLC(High Level Data Link Control)高级数据链路控制WLAN(Wireless Local Area Networks) 无线局域网MTBF 平均故障间隔时间DTE 数据终端设备DCE 数据通信设备TTL(Time To Time) 存在时间RAID(Redundant Arrays of Inexpensive Disk) 廉价磁盘冗余阵列NAS(Network Attached Storage) 网络连接存储SAN(Storage Area Network) 存储区域网络CGMP 分组管理协议Multicast 组播PDC 主域控制器RARP(Reverse Address Resolutuon Protocol) 逆向地址解析协议IGP(Interior Gateway Protocol) 内部网关协议EGP(Exterior Gateway Protocol) 外部网关协议NA T(Network Address Translation) 网络地址翻译CIDR(Classless Inter-Domain Routing) 无类别的域间路由技术HFC(Hybrid Fiber-Coax) 混合光纤同轴电缆ISDN(Integrated Service Digital Network) 综合业务数字网DQDB(Distributed Queue Dual Bus) 分布式队列双总线RMON(Remote Monitoring) 远程网络监控L2TP(Layer 2 Tunneling Protocol) 第而层通道协议DDR(Dial on Demand Routing) 按需拨号路由RIP(Routing Information Protocol) 路由选择信息协议STP(Spanning Tree Protocol) 生成树协议PPTP(Point to point Tunneling Protocol) 点对点隧道协议ESP(Encapsulating Security Payload) 封装安全负荷KMI(Key Management Infrastructure)密匙管理基础结构PKI(Public Key Infrastructure) 公匙基础结构SHA(The Secure Hash Alogorithm) 安全散列算法HMAC(Hashed Message Authencation Code) 散列式报文认证码CA(Certification Authority) 证书发放机构SSL(Secure Socket Layer) 安全套接层SET(Secure Electronic Transactoin) 安全的电子交易KDC(Key Distribution Center) 密匙分发中心ADSL(Asymmetrical Digital Subscriber Line) 非对称数字用户线扩展频谱通信:早期的扩频方式是频率跳动扩展频谱(Frequency Hopping Spread Spectrum,FHSS),更新的版本是直接序列扩展频谱(Direct Sequence Spread Spectrum,DSSS)第三层交换技术:1 IP交换IP交换机之间的信令使用了两个协议:IFMP(Ipsilon Flow Management Protocol,流管理协议)和GSMP(General Switch Management Protocol,交换机管理协议)2.MPLS多协议标记交换(Multi –Protocol Label Switching,MPLS),支持各种网络层协议,例如Ipv4 .Ipv6.IPX.CLNP等,同时MPLS特支持多种第二层协议.支持任何能够杂一网络层实体间传送分组的第二层煤体.而不是针对某一种链路技术.码分多址技术移动通讯系统有多种.按信号性质,有数字.模拟之分;按调制方式,又有调频.调相.调幅三类;按多址连接方式,又可分为频分多址(FDMA).时分多址(TDMA)和码分多址(CDMA)几种CDMA(Coad Division Multiple Access) 码分多址宽带无线接入技术主要有多通道多点分配业务(Multichannel Multipoint Dsitribution Services,MMDS)和本地多点分配业务(Local Multipoint Dstributon Services,LMDS)两种.他们在成熟的微波传输技术上发展起来的,所采用的调制方式与微波传输相似,主要为相移键控PSK(包括BPSK.DQPSK.QPSK.8PSK等)和正交幅度调制QAM(4-QAM.16-QAM.64-QAM等).不同之处是MMDS和LMDS均采用一点多址方式,微波传输采用点对点方式.幅度键控ASK频移键控FSK相移键控PSK1综合业务数字网(ISDN)用户设备分为两种类型:1型终端设备(TE1)符合ISDN接口标准,可以通过数字管道直接连接ISDN,例如数字电话,数字传真等;2型终端设备(TE2)是非标准的用户设备,必须通过终端适配器(TA)才能连接ISDN.通常的PC机就是TE2设备,需要插入一个ISDN适配卡才能接入ISDN.2能接入X.25 PDN网设备(DEE)可以分为两类:一类为分组终端(PT),另一类为非分组终端(NPT).分组终端是指具有X.25规程的所规定的所有功能的用户设备,这类设备可按X.25规程以同步方式接入X.25 PDN;非分组终端是指不具备X.25规程规定功能的用户设备.例如PC 机.为了对非分组设备提供接入接口,X.25 PDN定义了PAD(Packet Assemble Dissembler,分组拆装设备).PAD实际上是一个规程转换器,对非分组设备的规程进行转换.。

Network Working Group N. Kushalnagar Request for Comments: 4919 Intel Corp Category: Informational G. Montenegro Microsoft Corporation C. Schumacher Danfoss A/S August 2007 IPv6 over Low-Power Wireless Personal Area Networks (6LoWPANs):Overview, Assumptions, Problem Statement, and GoalsStatus of This MemoThis memo provides information for the Internet community. It doesnot specify an Internet standard of any kind. Distribution of thismemo is unlimited.Copyright NoticeCopyright (C) The IETF Trust (2007).AbstractThis document describes the assumptions, problem statement, and goals for transmitting IP over IEEE 802.15.4 networks. The set of goalsenumerated in this document form an initial set only.Table of Contents1. Introduction (2)2. Overview (2)3. Assumptions (3)4. Problems (4)4.1. IP Connectivity (4)4.2. Topologies (5)4.3. Limited Packet Size (6)4.4. Limited Configuration and Management (6)4.5. Service Discovery (6)4.6. Security (6)5. Goals (7)6. Security Considerations (9)7. Acknowledgements (10)8. References (10)8.1. Normative References (10)8.2. Informative References (10)Kushalnagar, et al. Informational [Page 1]1. IntroductionLow-power wireless personal area networks (LoWPANs) comprise devices that conform to the IEEE 802.15.4-2003 standard by the IEEE[IEEE802.15.4]. IEEE 802.15.4 devices are characterized by shortrange, low bit rate, low power, and low cost. Many of the devicesemploying IEEE 802.15.4 radios will be limited in their computational power, memory, and/or energy availability.This document gives an overview of LoWPANs and describes how theybenefit from IP and, in particular, IPv6 networking. It describesLoWPAN requirements with regards to the IP layer and the above, andspells out the underlying assumptions of IP for LoWPANs. Finally, it describes problems associated with enabling IP communication withdevices in a LoWPAN, and defines goals to address these in aprioritized manner. Admittedly, not all items on this list may benecessarily appropriate tasks for the IETF. Nevertheless, they aredocumented here to give a general overview of the larger problem.This is useful both to structure work within the IETF as well as tobetter understand how to coordinate with external organizations.2. OverviewA LoWPAN is a simple low cost communication network that allowswireless connectivity in applications with limited power and relaxed throughput requirements. A LoWPAN typically includes devices thatwork together to connect the physical environment to real-worldapplications, e.g., wireless sensors. LoWPANs conform to the IEEE802.15.4-2003 standard [IEEE802.15.4].Some of the characteristics of LoWPANs are as follows:1. Small packet size. Given that the maximum physical layer packet is 127 bytes, the resulting maximum frame size at the mediaaccess control layer is 102 octets. Link-layer security imposes further overhead, which in the maximum case (21 octets ofoverhead in the AES-CCM-128 case, versus 9 and 13 for AES-CCM-32 and AES-CCM-64, respectively), leaves 81 octets for datapackets.2. Support for both 16-bit short or IEEE 64-bit extended mediaaccess control addresses.3. Low bandwidth. Data rates of 250 kbps, 40 kbps, and 20 kbps for each of the currently defined physical layers (2.4 GHz, 915 MHz, and 868 MHz, respectively).4. Topologies include star and mesh operation.Kushalnagar, et al. Informational [Page 2]5. Low power. Typically, some or all devices are battery operated.6. Low cost. These devices are typically associated with sensors, switches, etc. This drives some of the other characteristicssuch as low processing, low memory, etc. Numerical values for"low" elided on purpose since costs tend to change over time.7. Large number of devices expected to be deployed during thelifetime of the technology. This number is expected to dwarfthe number of deployed personal computers, for example.8. Location of the devices is typically not predefined, as theytend to be deployed in an ad-hoc fashion. Furthermore,sometimes the location of these devices may not be easilyaccessible. Additionally, these devices may move to newlocations.9. Devices within LoWPANs tend to be unreliable due to variety ofreasons: uncertain radio connectivity, battery drain, devicelockups, physical tampering, etc.10. In many environments, devices connected to a LoWPAN may sleepfor long periods of time in order to conserve energy, and areunable to communicate during these sleep periods.The following sections take into account these characteristics indescribing the assumptions, problems statement, and goals forLoWPANs, and, in particular, for 6LoWPANs (IPv6-based LoWPANnetworks).3. AssumptionsGiven the small packet size of LoWPANs, this document presumesapplications typically send small amounts of data. However, theprotocols themselves do not restrict bulk data transfers.LoWPANs, as described in this document, are based on IEEE802.15.4-2003. It is possible that the specification may undergochanges in the future and may change some of the requirementsmentioned above.Some of these assumptions are based on the limited capabilities ofdevices within LoWPANs. As devices become more powerful, and consume less power, some of the requirements mentioned above may be somewhat relaxed.Kushalnagar, et al. Informational [Page 3]While some LoWPAN devices are expected to be extremely limited (theso-called "Reduced Function Devices" or RFDs), more capable "FullFunction Devices" (FFDs) will also be present, albeit in much smaller numbers. FFDs will typically have more resources and may be mainspowered. Accordingly, FFDs will aid RFDs by providing functions such as network coordination, packet forwarding, interfacing with othertypes of networks, etc.The application of IP technology is assumed to provide the following benefits:1. The pervasive nature of IP networks allows use of existinginfrastructure.2. IP-based technologies already exist, are well-known, and provento be working.3. An admittedly non-technical but important consideration is thatIP networking technology is specified in open and freelyavailable specifications, which is favorable or at least able to be better understood by a wider audience than proprietarysolutions.4. Tools for diagnostics, management, and commissioning of IPnetworks already exist.5. IP-based devices can be connected readily to other IP-basednetworks, without the need for intermediate entities liketranslation gateways or proxies.4. ProblemsBased on the characteristics defined in the overview section, thefollowing sections elaborate on the main problems with IP forLoWPANs.4.1. IP ConnectivityThe requirement for IP connectivity within a LoWPAN is driven by the following:1. The many devices in a LoWPAN make network auto configuration and statelessness highly desirable. And for this, IPv6 has readysolutions.2. The large number of devices poses the need for a large addressspace, well met by IPv6.Kushalnagar, et al. Informational [Page 4]3. Given the limited packet size of LoWPANs, the IPv6 address format allows subsuming of IEEE 802.15.4 addresses if so desired.4. Simple interconnectivity to other IP networks including theInternet.However, given the limited packet size, headers for IPv6 and layersabove must be compressed whenever possible.4.2. TopologiesLoWPANs must support various topologies including mesh and star.Mesh topologies imply multi-hop routing, to a desired destination.In this case, intermediate devices act as packet forwarders at thelink layer (akin to routers at the network layer). Typically theseare "full function devices" that have more capabilities in terms ofpower, computation, etc. The requirements on the routing protocolare:1. Given the minimal packet size of LoWPANs, the routing protocolmust impose low (or no) overhead on data packets, hopefullyindependently of the number of hops.2. The routing protocols should have low routing overhead (lowchattiness) balanced with topology changes and powerconservation.3. The computation and memory requirements in the routing protocolshould be minimal to satisfy the low cost and low powerobjectives. Thus, storage and maintenance of large routingtables is detrimental.4. Support for network topologies in which either FFDs or RFDs maybe battery or mains-powered. This implies the appropriateconsiderations for routing in the presence of sleeping nodes.As with mesh topologies, star topologies include provisioning asubset of devices with packet forwarding functionality. If, inaddition to IEEE 802.15.4, these devices use other kinds of networkinterfaces such as ethernet or IEEE 802.11, the goal is to seamlessly integrate the networks built over those different technologies.This, of course, is a primary motivation to use IP to begin with. Kushalnagar, et al. Informational [Page 5]4.3. Limited Packet SizeApplications within LoWPANs are expected to originate small packets. Adding all layers for IP connectivity should still allow transmission in one frame, without incurring excessive fragmentation andreassembly. Furthermore, protocols must be designed or chosen sothat the individual "control/protocol packets" fit within a single802.15.4 frame. Along these lines, IPv6’s requirement of sub-IPreassembly (see Section 5) may pose challenges for low-end LoWPANdevices that do not have enough RAM or storage for a 1280-octetpacket.4.4. Limited Configuration and ManagementAs alluded to above, devices within LoWPANs are expected to bedeployed in exceedingly large numbers. Additionally, they areexpected to have limited display and input capabilities.Furthermore, the location of some of these devices may be hard toreach. Accordingly, protocols used in LoWPANs should have minimalconfiguration, preferably work "out of the box", be easy tobootstrap, and enable the network to self heal given the inherentunreliable characteristic of these devices. The size constraints of the link layer protocol should also be considered. Networkmanagement should have little overhead, yet be powerful enough tocontrol dense deployment of devices.4.5. Service DiscoveryLoWPANs require simple service discovery network protocols todiscover, control and maintain services provided by devices. In some cases, especially in dense deployments, abstraction of several nodes to provide a service may be beneficial. In order to enable suchfeatures, new protocols may have to be designed.4.6. SecurityIEEE 802.15.4 mandates link-layer security based on AES, but it omits any details about topics like bootstrapping, key management, andsecurity at higher layers. Of course, a complete security solutionfor LoWPAN devices must consider application needs very carefully.Please refer to the security consideration section below for a moredetailed discussion and in-depth security requirements.Kushalnagar, et al. Informational [Page 6]5. GoalsThe goals mentioned below are general and not limited to IETFactivities. As such, they may not only refer to work that can bedone within the IETF (e.g., specification required to transmit IP,profile of best practices for transmitting IP packets, and associated upper level protocols, etc). They also point at work more relevantto other standards bodies (e.g., desirable changes to or profilesrelevant to IEEE 802.15.4, W3C, etc). When the goals fall under the IETF’s purview, they serve to point out what those efforts shouldstrive to accomplish, regardless of whether they are pursued withinone (or more) new (or existing) working groups. When the goals donot fall under the purview of the IETF, documenting them here serves as input to other organizations [LIAISON].Note that a common underlying goal is to reduce packet overhead,bandwidth consumption, processing requirements, and powerconsumption.The following are the goals according to priority for LoWPANs:1. Fragmentation and Reassembly layer: As mentioned in the overview, the protocol data units may be as small as 81 bytes. This isobviously far below the minimum IPv6 packet size of 1280 octets, and in keeping with Section 5 of the IPv6 specification[RFC2460], a fragmentation and reassembly adaptation layer mustbe provided at the layer below IP.2. Header Compression: Given that in the worst case the maximum size available for transmitting IP packets over an IEEE 802.15.4 frame is 81 octets, and that the IPv6 header is 40 octets long,(without optional headers), this leaves only 41 octets forupper-layer protocols, like UDP and TCP. UDP uses 8 octets inthe header and TCP uses 20 octets. This leaves 33 octets fordata over UDP and 21 octets for data over TCP. Additionally, as pointed above, there is also a need for a fragmentation andreassembly layer, which will use even more octets leaving veryfew octets for data. Thus, if one were to use the protocols asis, it would lead to excessive fragmentation and reassembly, even when data packets are just 10s of octets long. This points tothe need for header compression. As there is much published and in-progress standardization work on header compression, the6LoWPAN community needs to investigate using existing headercompression techniques, and, if necessary, specify new ones. Kushalnagar, et al. Informational [Page 7]3. Address Autoconfiguration: [6LoWPAN] specifies methods forcreating IPv6 stateless address auto configuration. Statelessauto configuration (as compared to stateful) is attractive for6LoWPANs, because it reduces the configuration overhead on thehosts. There is a need for a method to generate an "interfaceidentifier" from the EUI-64 [EUI64] assigned to the IEEE 802.15.4 device.4. Mesh Routing Protocol: A routing protocol to support a multi-hop mesh network is necessary. There is much published work on ad-hoc multi hop routing for devices. Some examples include[RFC3561], [RFC3626], [RFC3684], all experimental. Also, theseprotocols are designed to use IP-based addresses that have large overheads. For example, the Ad hoc On-Demand Distance Vector(AODV) [RFC3561] routing protocol uses 48 octets for a routerequest based on IPv6 addressing. Given the packet-sizeconstraints, transmitting this packet without fragmentation andreassembly may be difficult. Thus, care should be taken whenusing existing routing protocols (or designing new ones) so that the routing packets fit within a single IEEE 802.15.4 frame.5. Network Management: One of the points of transmitting IPv6packets is to reuse existing protocols as much as possible.Network management functionality is critical for LoWPANs.However, management solutions need to meet the resourceconstraints as well as the minimal configuration and self-healing functionality described in Section 4.4. The Simple NetworkManagement Protocol (SNMP) [RFC3410] is widely used formonitoring data sources and sensors in conventional networks.SNMP functionality may be translated "as is" to LoWPANs with the benefit to utilize existing tools. However, due to the memory,processing, and message size constraints, further investigationis required to determine if the use of SNMPv3 is suitable, or if an appropriate adaptation of SNMPv3 or use of different protocols is in order.6. Implementation Considerations: It may be the case thattransmitting IP over IEEE 802.15.4 would become more beneficialif implemented in a "certain" way. Accordingly, implementationconsiderations are to be documented.7. Application and higher layer Considerations: As headercompression becomes more prevalent, overall performance willdepend even more on efficiency of application protocols.Heavyweight protocols based on XML such as SOAP [SOAP], may notbe suitable for LoWPANs. As such, more compact encodings (andperhaps protocols) may become necessary. The goal here is tospecify or suggest modifications to existing protocols so that Kushalnagar, et al. Informational [Page 8]they are suitable for LoWPANs. Furthermore, application levelinteroperability specifications may also become necessary in the future and may thus be specified.8. Security Considerations: Security threats at different layersmust be clearly understood and documented. Bootstrapping ofdevices into a secure network could also be considered given the location, limited display, high density, and ad-hoc deployment of devices.6. Security ConsiderationsIPv6 over LoWPAN (6LoWPAN) applications often require confidentiality and integrity protection. This can be provided at the application,transport, network, and/or at the link layer (i.e., within the6LoWPAN set of specifications). In all these cases, prevailingconstraints will influence the choice of a particular protocol. Some of the more relevant constraints are small code size, low poweroperation, low complexity, and small bandwidth requirements.Given these constraints, first, a threat model for 6LoWPAN devicesneeds to be developed in order to weigh any risks against the cost of their mitigations while making meaningful assumptions andsimplifications. Some examples for threats that should be considered are man-in-the-middle attacks and denial of service attacks.A separate set of security considerations apply to bootstrapping a6LoWPAN device into the network (e.g., for initial keyestablishment). This generally involves application level exchanges or out-of-band techniques for the initial key establishment, and may rely on application-specific trust models; thus, it is consideredextraneous to 6LoWPAN and is not addressed in these specifications.In order to be able to select (or design) this next set of protocols, there needs to be a common model of the keying material created bythe initial key establishment.Beyond initial key establishment, protocols for subsequent keymanagement as well as to secure the data traffic do fall under thepurview of 6LoWPAN. Here, the different alternatives (TLS, IKE/IPsec, etc.) must be evaluated in light of the 6LoWPAN constraints.One argument for using link layer security is that most IEEE 802.15.4 devices already have support for AES link-layer security. AES is ablock cipher operating on blocks of fixed length, i.e., 128 bits. To encrypt longer messages, several modes of operation may be used. The earliest modes described, such as ECB, CBC, OFB and CFB provide only confidentiality, and this does not ensure message integrity. Othermodes have been designed which ensure both confidentiality and Kushalnagar, et al. Informational [Page 9]message integrity, such as CCM* mode. 6LoWPAN networks can operate in any of the previous modes, but it is desirable to utilize the mostsecure modes available for link-layer security (e.g., CCM*), andbuild upon it.For network layer security, two models are applicable: end-to-endsecurity, e.g., using IPsec transport mode, or security that islimited to the wireless portion of the network, e.g., using asecurity gateway and IPsec tunnel mode. The disadvantage of thelatter is the larger header size, which is significant at the 6LoWPAN frame MTUs. To simplify 6LoWPAN implementations, it is beneficial to identify the relevant security model, and to identify a preferred set of cipher suites that are appropriate given the constraints.7. AcknowledgementsThanks to Geoff Mulligan, Soohong Daniel Park, Samita Chakrabarti,Brijesh Kumar, and Miguel Garcia for their comments and help inshaping this document.8. References8.1. Normative References[RFC2460] Deering, S. and R. Hinden, "Internet Protocol, Version 6 (IPv6) Specification", RFC 2460, December 1998.[IEEE802.15.4] IEEE Computer Society, "IEEE Std. 802.15.4-2003",October 2003.8.2. Informative References[EUI64] "GUIDELINES FOR 64-BIT GLOBAL IDENTIFIER (EUI-64)REGISTRATION AUTHORITY", IEEE,/regauth/oui/tutorials/EUI64.html.[6LoWPAN] Thomson, S., Narten, T., and T. Jinmei, "IPv6Stateless Address Autoconfiguration", Work inProgress, May 2005.[RFC3411] Harrington, D., Presuhn, R., and B. Wijnen, "AnArchitecture for Describing Simple Network Management Protocol (SNMP) Management Frameworks", STD 62, RFC3411, December 2002.Kushalnagar, et al. Informational [Page 10][RFC3561] Perkins, C., Belding-Royer, E., and S. Das, "Ad hocOn-Demand Distance Vector (AODV) Routing", RFC 3561,July 2003.[RFC3626] Clausen, T. and P. Jacquet, "Optimized Link StateRouting Protocol (OLSR)", RFC 3626, October 2003.[RFC3684] Ogier, R., Templin, F., and M. Lewis, "TopologyDissemination Based on Reverse-Path Forwarding(TBRPF)", RFC 3684, February 2004.[SOAP] "XML Protocol Working Group", W3C,/2000/xp/Group/.[LIAISON] "IETF Liaison Activities", IETF,/liaisonActivities.html.Authors’ AddressesNandakishore KushalnagarIntel CorpEMail: nandakishore.kushalnagar@Gabriel MontenegroMicrosoft CorporationEMail: gabriel.montenegro@Christian Peter Pii SchumacherDanfoss A/SEMail: schumacher@Kushalnagar, et al. Informational [Page 11]Full Copyright StatementCopyright (C) The IETF Trust (2007).This document is subject to the rights, licenses and restrictionscontained in BCP 78, and except as set forth therein, the authorsretain all their rights.This document and the information contained herein are provided on an "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY, THE IETF TRUST AND THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIEDWARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Intellectual PropertyThe IETF takes no position regarding the validity or scope of anyIntellectual Property Rights or other rights that might be claimed to pertain to the implementation or use of the technology described inthis document or the extent to which any license under such rightsmight or might not be available; nor does it represent that it hasmade any independent effort to identify any such rights. Information on the procedures with respect to rights in RFC documents can befound in BCP 78 and BCP 79.Copies of IPR disclosures made to the IETF Secretariat and anyassurances of licenses to be made available, or the result of anattempt made to obtain a general license or permission for the use of such proprietary rights by implementers or users of thisspecification can be obtained from the IETF on-line IPR repository at /ipr.The IETF invites any interested party to bring to its attention anycopyrights, patents or patent applications, or other proprietaryrights that may cover technology that may be required to implementthis standard. Please address the information to the IETF atietf-ipr@.AcknowledgementFunding for the RFC Editor function is currently provided by theInternet Society.Kushalnagar, et al. Informational [Page 12]。

数据中心存储网络：数据存储已经成为目前的一个热点技术，也是继互联网热之后的又一次技术浪潮，它将网络带入以数据为中心的时代。

数据存储经过了三个发展阶段：直接附加存储（DAS）、网络附加存储(NAS) 、存储区域网络（SAN）。

DAS是在以CPU为中心的计算为王时代的产物，适应于最初计算机工业的发展，而对于SAN和NAS，其技术上最大的区别在于是采用专门的协议还是现有的IP技术，以及数据共享等问题的分别考虑SAN的优势在于最初解决网络带宽问题的考虑NAS更侧重于通用性和数据共享的考虑。

存储分类：DAS：Direct-Attached StorageFAS：Fabric-Attached StorageNAS：Network Attached StorageSAN：Storage Networks存储模型的比较：DAS（Direct Attached Storage 直接附加存储）是指将存储设备通过SCSI线缆或光纤通道直接连接到服务器上SAN（Storage Area Network 存储区域网络），是一种通过网络方式连接存储设备和应用服务器的存储构架，这个网络专用于主机和存储设备之间的访问，存储设备串行运行，文件共享程度低NAS（Network Attached Storage 网络附加存储），是一种文件共享服务。

存储系统拥有自己的文件系统，通过NFS或CIFS（SMB）对外提供文件访问服务SAN与NAS：NAS（Network- Attached Storage ）协议:NFS/CIFS (基于TCP/IP)提供文件访问，适用于文件存储需求适宜文件共享访问的应用，支持异构平台文件共享文件的数据迁移比裸设备简单可靠SAN（Storage Area Networks）协议: FC/iSCSI裸设备访问，适宜传统数据库访问依赖应用主机提供文件访问。

共享访问需要集群软件支持，处理冲突访问开销大，性能较差，难以支持异构环境共享裸设备数据迁移困难DAS、SAN、NAS对比：DAS - 传统直连存储技术：适合于对存储容量要求不高、服务器的数量很少的中小型局域网，其主要的优点在于存储容量扩展的实施非常简单，投入的成本少而见效快特点：存储设备直接连接到主机数据分散管理存储容量利用率低扩展性差DAS的定义：DAS是1个或多个直接连接到使用它们的服务器上的指定存储设备，这些存储设备为服务器提供块级数据访问服务DAS分类：基于存储设备与服务器间的位置关系，DAS分为内部DAS和外部DAS：内部DAS：在内部DAS架构中，存储设备通过服务器机箱内部的并行或串行总线连接到服务器上。

IBM Storage Networking SAN24B-6 swi tchUse a low-cost, high-perfor mance platform to enable flash-r eady solutions for the businessHighlightsMeet the high-thr oughput, low-la tency demands of critical applications with flash-ready performanceScale on demand, from 8 to 24 ports, to connect additional devices as neededDeliver 4, 8, 16 or 32 Gigabits per second (Gbps) port bandwidth for increased performance on demandSimplify deployment and reduce install time with a point-a nd- c lick user interface Automatically discover and recover from common networking problemsProactively monitor and optimisethe health and performance of individual VMsSimplify administration, resolve problems, increase uptime and help reduce costs using Fabric Vision technology.Explosive data growth, coupled with user expectations of unlimited access from anywhere, at any time, is pushing storage environmentsto the limit. T o meet these dynamic business demands, the network must evolve to improve speed, increase efficiency and reduce costs. Legacy infrastructures were not designed to support the performance requirements of flash- b ased storage technology. A new approachto storage networking is required to unlock the full capabilities ofall-f lash arrays. By treating the network as a strategic part of a storage environment, organisations can maximise their productivity and efficiency, even as they rapidly grow their environments.The IBM Storage Networking SAN24B-6 switch provides exceptional value in an entry-level switch, combining high-performance capabilities of 4, 8, 16 and 32 Gbps, point-and-click simplicity and enterprise-class functionality. The port speed capability is dependent on the transceiver installed.SAN24B-6 provides small to midsized data centres (DCs) with low-cost access to leading-edge Gen 6 FC technology and the ability to start small and grow on demand – from 8 to 24 ports – to support an evolving storage environment. In addition, SAN24B-6 is easy to use and install, with a point-and-click user interface that simplifies deployment and saves time.Gen 6 Fibre ChannelIBM® b- t ype Gen 6 Fibre Channel (FC) is the purpose-built n etwork infrastructure for mission-cr itical storage, delivering breakthrough performance to accelerate data retrieval, adapt to evolving requirements and drive always-on da ta access. The SAN24B-6 s witch with Gen 6 FC storage technology is designed to simplify setup, management and monitoring while delivering performance required by flash storage and growing with your business at a low costof entry.Gain maximum flexibilityThe SAN24B-6 switch is configurable with Ports on Demand (PoD) for 8, 16 or 24 ports and supports 4, 8, 16 or 32 Gbps port speeds with the proper transceivers installed, all in an efficient 1U package, with an integrated power supply and four built-in fans. The power supply offers real-time, active power monitoring. SAN24B-6 helps to lower costs, too, with low energy consumption (at 0.10 watts per Gbps and 3.2 watts per port).Leading- e dge technology that is flexible, simple and easy to useSAN24B-6 delivers leading-edge Gen 6 FC technology in a flexible and easy-to-use solution that cost-effectively scales from 8 to 24 ports with PoD. In addition, the switch is easyto deploy with the EZSwitchSetup wizard, featuring a simple user interface that dramatically reduces deployment and configuration times with as few as three steps.Flash- r eady performance for evolving storage requirementsIBM b-type Gen 6 FC products deliver advanced 32-Gbps performance to unleash the full potential of new storage technologies for the new high-performance application workloads. Using this switch, organisations can build aflash-ready infrastructure that adapts to their expanding business requirements.T o realise the full benefits of flash, organisations will need to transition their high-performance, latency-sensitive workloads to flash- b ased storage with NVMe. SAN24B-6 is NVMe-ready, allowing organisations to seamlessly integrate IBM b-typeGen 6 FC networks with next-generation NVMe technologies, without a disruptive rip and replace. The simplicity and efficiency of NVMe over FC enables significant performance gains for flash storage to deliver the performance, application response time and scalability needed for next-generation DCs. Fabric Vision technologyFabric Vision technology with virtual machine (VM) Insightis an optional feature that provides unprecedented insight and visibility across the storage network to simplify monitoring, increase operational stability and dramatically reduce maintenance and diagnostic costs. For more details about Fabric Vision capabilities, review the information in the Fabric Vision data sheet1 and IBM Redbooks product guide. Figure 1. The IBM Storage Networking SAN24B- 6switch.Rely on the network that delivers always- o n business operationsIBM b-type Gen 5 and Gen 6 technologies leverage a rich heritage of FC innovation to deliver leading-edge reliability for the world’s most demanding DCs. IBM b-type Fabric V ision technology provides a breakthrough hardware and software solution that helps organisations simplify monitoring, maximise network availability and gain insight into issues to speed resolution and meet critical service level agreements (SLAs). VM Insight is the newest feature in Fabric Vision technology, enabling proactive visibility into the health and performance of individual VMs with integrated sensors. Using this capability, administrators can quickly identify abnormal VM behaviours to facilitate troubleshooting and fault isolation, helping to ensure maximum performance and operational stability.Forward error correction (FEC) capabilities further increase resiliency by automatically detecting and recovering network transmission errors. T o ensure predictable performance prior to deployment, organisations can validate infrastructure with ClearLink Diagnostics and Flow Generator features. Simplified management and robust network analyticsFabric Vision technology provides unprecedented insight and visibility across the storage network, with powerful integrated monitoring, management and diagnostic capabilities. These innovative features enable administrators to avoid problems before they impact operations, helping organisations meet SLAs. Fabric Vision technology includes VM Insight, Monitoring and Alerting Policy Suite (MAPS), Fabric Performance Impact (FPI) Monitoring, dashboards, Configuration and Operational Monitoring Policy Automation Services Suite (COMPASS), ClearLink Diagnostics, Flow Vision, FEC and Credit Loss Recovery. For additional information about the aforementioned capabilities, please review the Fabric Vision technologyIBM Redbooks product guide.2IBM Network AdvisorIBM Network Advisor is an optional software management tool that provides an enhanced user interface and additional capabilities to simplify b-type FC management, helping organisations proactively diagnose and resolve issues to maximise uptime, increase operational efficiency and reduce costs. The wizard-driven interface dramatically reduces deployment and configuration times by allowing fabrics, switches and ports to be managed as groups. Customisable dashboards graphically display performance and health indicators out of the box, including all data captured using Fabric Vision technology. For further information, please refer to the IBM Network Advisor datasheet3 and IBM Redbooks product guide.4Access Gateway modeSAN24B-6 can be deployed as a full-fabric switch or as an Access Gateway, which simplifies fabric topologies and heterogeneous fabric connectivity (the default mode setting is a switch). Access Gateway mode utilises N_Port ID virtualisation (NPIV) switch standards to present physical and virtual servers directly to the core of storage area network (SAN) fabrics. This makes a switch in Access Gateway mode transparent to the SAN fabric, greatly reducing management of the overall storage network. SAN24B-6 in Access Gateway mode can connect servers to SAN fabrics that are NPIV-enabled. Maximising investmentsT o help optimise technology investments, IBM and its partners offer complete solutions that include professional services, technical support and education.IBM Storage Networking SAN24B- 6 switch at a glance Product number8960-F24Hot-swapcomponentsSmall form- f actor pluggables (SFPs)Warranty One- y ear; customer- r eplaceable unit (CRU) andon-site, 9×5 next-business-day response;warranty service upgrades are availablePort speed4, 8 and 16 Gbps when using 16 Gbps smallform- f actor pluggable (SFP+) transceivers8, 16 and 32 Gbps when using 32 Gbps SFP+transceiversOptional features Please refer to the SAN24B- 6 Redbooks Product Guide to review most current optional featuresSize Width: 42.88 cm (16.88 in.)Height: 4.29 cm (1.69 in.)Depth: 30.66 cm (12.07 in.)Weight 5.76 kg (12.65 lb) with one integrated powersupply, without transceiversWhy IBM?High-speed optical storage networking has long been out of reach for small and midsized enterprises. IBM can help growing businesses with an affordable, flexible, flash-ready switch solution. The new IBM Storage Networking SAN24B-6 switch delivers uncompromising speed for critical environments, with capabilities that can be expanded with the innovative POD capability.For more informationT o learn more about the IBM Storage Networking SAN24B-6 switch, please contact your IBM representative or IBM Business Partner (BP), or visit: /us-en/marketplace/san24b-6 Additionally, IBM Global Financing provides numerous payment options to help you acquire the technology you need to grow your business. We provide full lifecycle management of IT products and services, from acquisition to disposition. For more information, visit: /financing IBM United Kingdom LimitedPO Box 41North HarbourPortsmouthHampshirePO6 3AUUnited KingdomIBM Ireland LimitedOldbrook House24-32 Pembroke RoadDublin 4IBM Ireland Limited registered in Ireland under company number 16226. The IBM home page can be found at IBM, the IBM logo, and Redbooks are trademarks or registered trademarks of International Business Machines Corporation in the United States, other countries, or both. If these and other IBM trademarked terms are marked on their first occurrence in this information with a trademark symbol (® or ™), these symbols indicate U.S. registered or common law trademarks owned by IBM at the time this information was published. Such trademarks may also be registered or common law trademarks in other countries.A current list of IBM trademarks is available on the Web at ‘Copyright and trademark information’ at /leg al/copytrade.shtmlOther company, product and service names may be trademarks, or service marks of others.References in this publication to IBM products, programs or services do not imply that IBM intends to make these available in all countries in which IBM operates.Any reference to an IBM product, program or service is not intended to imply that only IBM products, programs or services may be used. Any functionally equivalent product, program or service may be used instead. IBM hardware products are manufactured from new parts, or new and used parts. In some cases, the hardware product may not be new and may have been previously installed. Regardless, IBM warranty terms apply. This publication is for general guidance only.Information is subject to change without notice. Please contact your local IBM sales office or reseller for latest information on IBM products and services.This publication contains non-IBM Internet addresses. IBM is not responsible for information found at these Web sites.IBM does not provide legal, accounting or audit advice or representor warrant that its products or services ensure compliance with laws. Clients are responsible for compliance with applicable securities lawsand regulations, including national laws and regulations.Photographs may show design models.© Copyright IBM Corporation 20171Based on Brocade Global Support analysis of customer support issuesthat have been escalated to Brocade.2Based on Brocade analysis of typical maintenance costs.3Based on a price comparison against competitors with tool forauto-registered phones support (T APS) to provide monitoring.4Access Gateway mode for SAN24B-6 is supported only in 24-portconfigurations.Please Recycle。

NAS存储技术资料大整合1. 概述近年来，服务器CPU的速度在不断翻番，随着ATM和千兆以太网的普遍使用，网络带宽不足的矛盾也得到很大缓解。

硬盘系统的容量在不断扩大，但I/0的速度始终没有明显的提高，这就使它日益明显地成为制约整个网络系统效率的瓶颈。

网上应用的数据量在迅速膨胀，使数据管理变得越来越复杂。

为了简化网络数据管理的复杂性，近年来出现了存贮网络技术。

它将硬盘系统与单个的应用服务器脱离，直接挂在网上，将分散的数据集中管理。

随着网络应用的普及，网络系统的可靠性日益成为关系企业生存的关键性因素。

为此在存贮系统中越来越广泛地采用磁盘阵列（RAID）技术。

但通常的RAID系统是用牺牲效率换取可靠性，这就使存贮系统效率的瓶颈问题更加突出。

NAS系统直接挂在网上的专用文件服务器，具备快速，简单，可靠的性能，支持Unix 和WindowNT多种网络环境。

2. 背景随着Internet的发展，数据成爆炸型增长，而现有网络环境却没有为它提供充分的发展空间。

网络系统每时每刻都在调用传输处于服务器后端磁盘阵列中的数据。

在很多大的应用系统中，硬盘I/O占用了服务器的大量时间，整个系统都处于对硬盘的等待状态，使整个系统性能下降。

磁盘成了服务器的包袱，制约整个网络的瓶颈。

回顾构成整个网络环境的几个关键部件的发展历程，CPU经历了由80286、80386、……、AlphaGHz等过程，可以说目前CPU技术为当前提供了一个比较好的计算环境，而网络带宽由10M、100M发展到目前技术成熟的1000M以太网为当前网络发展提供了足够的带宽，而硬盘的容量也在飞速发展，目前一个硬盘的容量25GB也是常事，但是从始至终一直有一个制约硬盘发展的问题没有得到很好的解决，这就是磁头臂机械移动速度一直没有得到提高，但磁头臂的机械移动速度却已经达到了一个极限。

硬盘就好比一个瓶子，瓶肚越来越大，但瓶口却是一直没有改变。

回顾整个计算机发展历程，不难发现计算机走了一条由通用→专用的道路。

目录一、磁盘I/O 的概念 (1)二、性能评价指标 (2)IOPS 与吞吐量的关系 (2)三、I/O 读写的类型 (5)大/ 小块I/O (5)连续/ 随机I/O (5)顺序/ 并发I/O (5)四、磁盘I/O 性能的监控 (6)topas (6)nmon (7)五、磁盘I/O 性能调优 (7)确认磁盘I/O 存在性能问题 (7)一、磁盘 I/O 的概念I/O 的概念，从字义来理解就是输入输出。

操作系统从上层到底层，各个层次之间均存在I/O。

比如，CPU 有I/O，内存有I/O, VMM 有I/O, 底层磁盘上也有I/O，这是广义上的I/O。

通常来讲，一个上层的I/O 可能会产生针对磁盘的多个I/O，也就是说，上层的I/O 是稀疏的，下层的I/O 是密集的。

磁盘的I/O，顾名思义就是磁盘的输入输出。

输入指的是对磁盘写入数据，输出指的是从磁盘读出数据。

我们常见的磁盘类型有ATA、SATA、FC、SCSI、SAS，如图1所示。

这几种磁盘中，服务器常用的是SAS 和FC 磁盘，一些高端存储也使用SSD 盘。

每一种磁盘的性能是不一样的。

图 1. 物理磁盘的架构以及常见磁盘类型二、性能评价指标SAN（Storage Area Network, 存储区域网络）和NAS存储（Network Attached Storage，网络附加存储)一般都具备2个评价指标：IOPS和带宽（throughput），两个指标互相独立又相互关联。

体现存储系统性能的最主要指标是IOPS。

下面，将介绍一下这两个参数的含义。

IOPS (Input/Output Per Second)即每秒的输入输出量(或读写次数)，是衡量磁盘性能的主要指标之一。

IOPS 是指单位时间内系统能处理的I/O请求数量，I/O请求通常为读或写数据操作请求。

随机读写频繁的应用，如OLTP(Online Transaction Processing)，IOPS是关键衡量指标。

竭诚为您提供优质文档/双击可除san,协议篇一：das、san、nas存储协议的工作原理目前磁盘存储市场上，存储分类（如下表一）根据服务器类型分为：封闭系统的存储和开放系统的存储，封闭系统主要指大型机，as400等服务器，开放系统指基于包括windows、unix、linux等操作系统的服务器；开放系统的存储分为：内置存储和外挂存储；开放系统的外挂存储根据连接的方式分为：直连式存储（direct-attachedstorage，简称das）和网络化存储（Fabric-attachedstorage，简称Fas）；开放系统的网络化存储根据传输协议又分为：网络接入存储（network-attachedstorage，简称nas）和存储区域网络（storageareanetwork，简称san）。

由于目前绝大部分用户采用的是开放系统，其外挂存储占有目前磁盘存储市场的70%以上，因此本文主要针对开放系统的外挂存储进行论述说明。

第一个图有问题，把nas和san一样放在Fas之下是不对的，通常也没有Fas这种说法，das，nas和san是平行的关系。

nas不一定要用光纤。

nas是文件级存储，san和das通常是数据块级存储。

表一：今天的存储解决方案主要为：直连式存储（das）、存储区域网络（san）、网络接入存储（nas）。

如下表二：开放系统的直连式存储（direct-attachedstorage，简称das）已经有近四十年的使用历史，随着用户数据的不断增长，尤其是数百gb以上时，其在备份、恢复、扩展、灾备等方面的问题变得日益困扰系统管理员。

主要问题和不足为：直连式存储依赖服务器主机操作系统进行数据的io读写和存储维护管理，数据备份和恢复要求占用服务器主机资源（包括cpu、系统io等），数据流需要回流主机再到服务器连接着的磁带机（库），数据备份通常占用服务器主机资源20-30%，因此许多企业用户的日常数据备份常常在深夜或业务系统不繁忙时进行，以免影响正常业务系统的运行。

磁盘存储DAS、NAS、SAN三种模式详解目前磁盘存储市场上，存储分类（如下表一）根据服务器类型分为：封闭系统的存储和开放系统的存储，封闭系统主要指大型机，AS400等服务器，开放系统指基于包括Windows、UNIX、Linux等操作系统的服务器；开放系统的存储分为：内置存储和外挂存储；开放系统的外挂存储根据连接的方式分为：直连式存储（Direct-Attached Storage，简称DAS）和网络化存储（Fabric-AttachedStorage，简称FAS）；开放系统的网络化存储根据传输协议又分为：网络接入存储（Network-AttachedStorage，简称NAS）和存储区域网络（Storage AreaNetwork，简称SAN）。

由于目前绝大部分用户采用的是开放系统，其外挂存储占有目前磁盘存储市场的70%以上，因此本文主要针对开放系统的外挂存储进行论述说明。

今天的存储解决方案主要为：直连式存储（DAS）、存储区域网络（SAN）、网络接入存储（NAS）。

如下表二：开放系统的直连式存储（Direct-Attached Storage，简称DAS）已经有近四十年的使用历史，随着用户数据的不断增长，尤其是数百GB以上时，其在备份、恢复、扩展、灾备等方面的问题变得日益困扰系统管理员。

主要问题和不足为：直连式存储依赖服务器主机操作系统进行数据的IO读写和存储维护管理，数据备份和恢复要求占用服务器主机资源（包括CPU、系统IO等），数据流需要回流主机再到服务器连接着的磁带机（库），数据备份通常占用服务器主机资源20-30%，因此许多企业用户的日常数据备份常常在深夜或业务系统不繁忙时进行，以免影响正常业务系统的运行。

直连式存储的数据量越大，备份和恢复的时间就越长，对服务器硬件的依赖性和影响就越大。

直连式存储与服务器主机之间的连接通道通常采用SCSI连接，带宽为10MB/s、20MB/s、40MB/s、80MB/s等，随着服务器CPU的处理能力越来越强，存储硬盘空间越来越大，阵列的硬盘数量越来越多，SCSI通道将会成为IO瓶颈；服务器主机SCSIID资源有限，能够建立的SCSI通道连接有限。