Network Performance Monitoring for Applications Using EXPAND.
Bj?rn Landfeldt*, Aruna Seneviratne*, Bob Melander**, Per Gunningberg**
*Dept. of Electrical Engineering and Telecommunications
The University of New South Wales
Sydney 2052 Australia
bjornl@https://www.doczj.com/doc/095899899.html,.au
a.seneviratne@https://www.doczj.com/doc/095899899.html,.au
**Information Technology, Dept. of Computer Systems
Uppsala University
P.O. Box 325, S-75105
Uppsala Sweden
[melander, perg]@docs.uu.se
Keywords: Network monitoring, wireless links, dynamic window generation, passive monitoring and active probing
Abstract:
As computing becomes more interactive and as networked applications need special system support, it has become increasingly important to manage network resources. An important part of the management lies within monitoring network behaviour. It is important for many applications to be able to estimate system performance in order to select appropriate encoding schemes, adaptation, buffering sizes etc. This paper discusses monitoring within wired and wireless networks and the type of monitoring information needed to support different applications. We suggest a hybrid active/passive monitoring approach with a dynamic time window mechanism and interchangeable filters to extract requested information. The paper also shows our initial experimental results and presents our conclusions.
1Introduction
A range of new applications for use with the Internet has been introduced in the past few years. Many of these are interactive and use data with real time characteristics. This has lead to a diversification in the way the Internet is used, and therefore changed the requirements of the network.
In parallel, there has been a rapid growth in cellular and other mobile communication systems such as wireless LANs, and infra red access nodes. Consequently, it is clear that the natural evolution is to combine the two and provide a wireless Internet, thus providing access to information “at anytime from anywhere”. The wireless Internet will comprise of a fixed IP based core network and multiple overlaid wireless access networks. This will provide users with simultaneous access to several different networks [14]. In the ideal case, this will provide the capability of matching the application and user requirements to network characteristics. For example, a user can interactively request a file download over a low latency cellular network, and let the bulk data transfer occur over a high speed satellite link.
In the wired Internet, resources can be reserved to satisfy the needs of applications. There are several existing technologies capable of doing this, such as ATM, Intserv [15] and Diffserv [16]. However, resource availability cannot be guaranteed in the wireless Internet since wireless links are unreliable by nature. The resource availability is not only governed by the maximum link speed and the traffic load, but is also dependant on external factors such as channel fading and noise interference. In this environment, it is important for applications to be able to determine the status of the network in order to adapt to the current conditions. For example, a video player can select encoding scheme according to the available bandwidth. It is therefore important for these applications to have access to a good monitoring tool to determine the characteristics of the network.
A number of network-monitoring tools have been proposed recently [2]. However, they have been primarily aimed at fixed Internet environments. Consequently, they are not suited for the emerging wireless Internet environment. In this paper, we discuss the requirements for performing network monitoring in wireless environments and suggest a framework that provides flexibility in statistical processing and presentation of monitoring information to enable the choice of network and the configuration of applications. The rest of this paper is organised as follows. Section 2 presents related work and section 3 discusses the requirements applications will have on a monitoring tool. Section 4 then discusses general aspects of network monitoring and section 5 examines the relevance of the proposed monitoring framework. Section 6 describes our experimental set-up and results and finally in section 7 we present our conclusions and future work.
2Related work
A good survey of network monitoring techniques and the available tools can be found in [2]. These monitoring schemes can be divided into two categories, namely active and passive monitoring.
2.1Active monitoring schemes
Active monitoring schemes operate by inserting probing traffic into the network. These probes are used to extract information about network characteristics such as available and bottleneck bandwidth, round trip delay etc. Bottleneck bandwidth is the bandwidth of the lowest bandwidth link of the entire transmission path when it is idle. Available bandwidth is the unused bandwidth taken over the path and it varies in time depending on cross traffic. The available bandwidth is always less than or equal to the bottleneck bandwidth.
Bottleneck bandwidth can be detected by a technique called back-to-back or packet-pair probing [6]. This technique measures the time space in-between two arriving packets that are sent back-to-back. If there are no other packets in-between the two probing packets, the space corresponds to the time it takes for the second packet to queue until the first packet is transmitted on the slowest link in the transmission path. Therefore, it is proportional to the bottleneck bandwidth. Several monitoring tools use this technique [5,6,7,11]. For example, Pathchar [7] uses this method in combination with traceroute [3], in order to derive per-hop information about the network.
The framework presented in [5] consists of two parts. The first part, b-probe, similarly to the other tools above, only provides information about the bottleneck bandwidth using packet-pair. The second part, c-probe, provides an estimate of the available bandwidth of a network. This is done by sending a “short train” (several packets send back-to-back) of ICMP echo packets and measuring the time in-between the first and last packet, that will eventually be received at the sender side.
A more common way of estimating the available bandwidth is to simulate a bulk data transfer. Ttcp [8] makes an estimation of the available bandwidth based on data from a TCP bulk data transfer. Treno [9] uses ICMP echo packets with flow and congestion control in order to simulate a TCP bulk data transfer, and from that derive the available bandwidth. Compared to probing that measures the distance between packets, both these schemes introduce significant traffic into the network. It should also be noted that these methods measure the average bandwidth over a period, and that the instantaneous bandwidth may vary significantly during this period.
The advantage of using the bottleneck bandwidth is that the value stays fixed as long as the routing path is not changed. The disadvantage is that the bottleneck value, though stable, does not reflect the actual load on a link and thus, may be an over-optimistic metric. Furthermore, bottleneck bandwidth estimation does not detect any degradation in throughput that might occur due to heavy CPU load in a server. We therefore believe that the available bandwidth is more useful for applications. Bottleneck bandwidth can be used for routing purposes, choice of network and for obtaining background statistical information that can be used for estimating bandwidth.
Active monitoring provides an accurate method of determining network characteristics. However, the monitoring traffic competes with application data flows for network resources and active probing is therefore not scalable, especially for resource scarce environments. Furthermore, the delays associated with obtaining necessary results may be large, thus making it unsuitable for use with reactive QoS managers such as USA. The manager reacts to user dissatisfaction and need to present configuration suggestions to the user within an “acceptable”period of time.
2.2Passive Monitoring
Passive schemes overcome the disadvantages of active schemes associated with overheads and delay by monitoring streams in progress and from them, deriving the current network status, rather than introducing monitoring traffic. SPAND [1] is a monitoring scheme that is based on the principles of passive monitoring. It extends the basic passive monitoring by providing facilities for sharing of measurement results among hosts in order to increase the accuracy. SPAND operates through a centralised server to which applications may report the measured parameters of a flow. This server is accessible by any application to obtain an estimate of the bandwidth to a certain host. The tool relies on the assumption that hosts within a certain region are likely to have the same path to a remote host and hence can share information. Furthermore,
it assumes that network performance is stable so that reported results at the server are valid for a “reasonable” period of time.
Although the last assumption is valid in the current fixed Internet environments as shown in [4],it is not valid in a wireless Internet environment. The performance of a wireless access network not only depends on the current load, but it also depends on other factors such as multi-path fading, other forms of interference, and the transmitted power. These factors in turn depend on the geographic location of the mobile host. Consequently, sharing measurement results collected by mobile hosts are likely to lead to unreliable estimates. Therefore, shared passive monitoring cannot be directly used in the wireless Internet environment.
3 Monitoring for applications
A network-monitoring tool should be versatile, in its ability to present information, thus supporting different applications as described earlier. Figure 1 shows an available bandwidth measurement, which illustrates the need for flexibility when presenting the results to the
application.
1
91725334149576573818997105113121129137145153161169177185193
T im e
B a n d w i d t h
Figure 1. Example of varying available bandwidth
The figure shows bandwidth varying with a “bounded noise” factor, the mean value of the
bandwidth and an effective “lower limit” which is the minimum measured bandwidth. Assume that the video conferencing tool in our example wants an estimate of the available bandwidth.The mean value might be a good estimate if the application is adaptive and can compensate for the noise, for example by using a play-out buffer. If not, the “lower limit” estimates the least available bandwidth the application will be expected to obtain from the network and the encoding scheme can be selected depending on this limit.
In order to highlight the need for versatility, table 1 gives a non-exhaustive list of the information different applications may require from a network-monitoring tool. A user running an FTP session may want to estimate the time of large file transfers and the associated monetary cost and battery usage. The mean available bandwidth is sufficient for this. A file transfer is elastic and therefore information about delay and jitter become irrelevant. In contrast, when using a web browser, the response time becomes one of the significant characteristics, therefore the monitoring should indicate the expected packet delay.
Table 1. Example of information required by different application types
As shown in table 1, stored video players need different information, e.g. mean available bandwidth for choosing the appropriate encoding and information about jitter in order to determine the play-out buffer size. Video applications may want to know the path Maximum Transmission Unit, MTU in order to be able to use Application Layer Framing, ALF [10].
4Hybrid Passive and Active Monitoring Framework
It is clear from the above discussion that neither active, nor passive monitoring can be directly used for monitoring a wireless Internet. However, it is possible to use a hybrid scheme that uses passive monitoring for the wired segments, and active monitoring for the wireless segments coupled with a dynamic windowing mechanism which accounts to the rapid fluctuations. The proposed scheme is such a hybrid scheme, which provides a universal monitoring tool for the emerging wireless Internet environment.
4.1Hybrid Monitoring
In the proposed scheme, as shown in Figure 2, the wired and wireless parts are treated independently. The passive component uses an extended SPAND server - EXPAND. The EXPAND server thus provides support for active monitoring of the wireless access networks as described below, in addition to the passive monitoring of the fixed network segment.
Figure 2. The hybrid passive and active monitoring approach
The hosts attached to the wireless segments will send active probe packets to the EXPAND server. These probe packets will request information from the EXPAND server about the connection via the wired segment to the destination. The EXPAND server will respond with the requested information. The request-response packet pair in turn will be used by the requesting host as active probes to determine the characteristics of the wireless access network. Finally, the requesting host will use the results form the EXPAND server about the fixed segment of the network, and the information about the wireless segment of the network, gathered from the request-response packet pair, to estimate the end-to-end characteristics of the connection.4.2 Dynamic Windowing
The raw measurement provided by the split-monitoring scheme in itself, is insufficient to obtain a good estimate of the network characteristics. Firstly, in order to obtain a good estimate the raw data needs to be filtered to suit the application requesting the data. Secondly, as the characteristics of the wireless access segment will vary significantly, it will not be possible to use simple statistical techniques. Therefore, the proposed scheme uses a dynamic windowing mechanism to filter the raw data obtained through the split-monitoring scheme. The dynamic windowing scheme exploits the fact that the Internet displays quasi-stable characteristics [4].This will be true even for the wireless networks as these networks will be used to access information from stationary locations, rather than whilst moving, i.e. from a meeting room, a laboratory or a client’s premises. The emerging wireless LAN environment contains mechanisms for reducing the data rate when the signal quality drops, thus giving different link speeds at different locations within the network [12, 13]. The quasi-stable behaviour also applies to cellular networks. Thus, within these networks, the data rate stays stable within a region and the only difference between the wireless and wired segments will be the period of stability.
Considering this, the proposed dynamic windowing scheme attempts to determine a step change in the network characteristics, and then uses a simple statistical analysis of the data from the step in order to determine the characteristics. This is schematically shown in Figure 3.
1
4
7
10
13
16
19
22
25
28
31
34
37
40
43
46
49
52
55
58
61
64
67
70
73
76
79
82
85
88
91
94
97
100
Time
B W
Figure 3. Dynamic window data regions
At time t = t x , 0 < t x < t 1, the data set used for calculation is between t x and t 0. At time t y ; t 1 < t y ,i.e. when the step change has been detected, the data set used for the calculation is between t y and t 1 and the data between t 0 and t 1 is discarded.
Considering the above, the windowing scheme consists of two parts. A mechanism for detecting step changes in instantaneous measured data and a statistical filtering mechanism. The step changes are determined by the standard deviation of the data normalised by the mean value. The standard deviation indicates the variability of the measurements and will therefore reflect the step change. The normalisation is done so that the variability can be quantified. In order to find the step, we define a threshold value δ, and compare it against the cumulative standard deviation to mean ratio, i.e.
A step change is found if ()
∑∑?==?≤
n i i
n
i x x x n i
n 112
111δ is true.
The statistical filtering mechanism then filters the data as described below and produces an
approximation of the requested parameters, i.e. RTT, bandwidth or jitter.
4.3 Filtering
The filtering phase should be versatile to support the need of different applications. USA will need a good estimate of the available bandwidth and delay. Other applications might need different information. For example, a video player might want to know jitter and mean available bandwidth values in order to determine the size of the play-out buffer. Interactive applications might want to know delay and an estimate of the periods with the lowest bandwidth for selecting an appropriate encoding scheme, since a play-out buffer cannot be used with an interactive application. Therefore, in the proposed scheme the filters are interchangeable, and specified by the requesting application as a part of the signalling while probing the wireless access network.The filters will approximate the data set within the region found by the dynamic windowing scheme. The region will typically consist of a trend and some noise. Since the traffic over the Internet is not yet fully modelled, it is impossible to determine which approximation is the most suitable. Initially we propose to approximate the data set with a curve, using the least square method in order to take advantage of the extra information the trend provides.
5 Experimental results
Figure 4. The experimental test-bed
To verify the viability of the above framework we conducted several tests within the experimental test-bed shown in Figure 4.
The WaveLAN network covered one floor of the Electrical Engineering building at the University of New South Wales, and provided varying SNR at different locations. The wireless tests consisted of a person moving with a laptop within the wireless LAN, measuring the available bandwidth to the PC from fixed locations. The wired tests were carried out over the Internet between the PC in Australia and the server in Sweden
5.1Wireless experiments
Our wireless tests aimed to investigate the traffic behaviour over the wireless link. Figure 5 shows a typical plot of available bandwidth measured over the wireless LAN while moving around. In the figure the client moved towards the cell boundary where the bandwidth dropped significantly. The client then turned and moved towards the centre again and the bandwidth rapidly increased to the maximum value. In our measurements, we could consistently see this behaviour.
Figure 5. Bandwidth measurements while moving in a wireless LAN.
Because of the stable behaviour within a cell and rapid changes at the boundaries, the need for a windowing scheme is accentuated. The stable behaviour also means historical data is useful even within a wireless network.
5.2Dynamic window experiments
We conducted a series of measurements between Australia and Sweden in order to investigate the viability of the dynamic window scheme. Figure 6 shows one measurement of the round trip time (RTT) measured between Australia and Sweden during 4 hours. The following example demonstrates the performance improvement of the dynamic window scheme.
As can be seen the measurements vary significantly with time. There is a significant change at around 8 am when most people come in to work and start to read mail, newspapers etc. The round trip time reaches a peak at around 8.30 am and starts to decay slowly until 9 am after which it drops down to its previous level.
The figure shows RTT estimations using three different methods, mean value, least square linear approximation and least square linear approximation using the dynamic window scheme. In order to find the edges of the dynamic windows we used the mean to standard deviation formula described above and compared it to a δ-value of 0.15. Determining the appropriate δ-value is still
an open research issue. For now, our experiences show that a value between 0.15 and 0.20 makes the scheme perform well.
0200
400
600
800
1000
1200
1400
1
12
23
34
45
56
67
78
89
100
111
122
133
144
155
166
177
188
199
210
221
232
243
254
265
276
287
298
309
320
331
342
353
Measurement
T i m e (m s )
Figure 6. RTT estimation between Australia and Sweden using three different methods
05000
10000
15000
20000
25000
3000035000
40000
45000
50000
12345678101112131415161718192122232425262728293032333435Measurement
E r r o r
Figure 7 Cumulative Error of estimated and actual RTT values
The δ-value of 0.15 results in the dynamic windowing scheme finding four different regions with values 1-21, 22-141, 142-241 and 242-359. The figure clearly shows that the least square linear approximation with dynamic windowing estimates the RTT better than the other two methods. In order to highlight the performance difference of the three methods, Figure 7 shows the cumulative error between estimated and actual RTT values. Again, the least square linear approximation with dynamic windowing performs much better than the other two methods.5.3 Filtering experiments
We have made preliminary experiments with different methods for filtering the data after we have applied the dynamic window scheme. From our measurements, we have found that the traffic over the fixed Internet show quasi-stable properties as stated previously. We have also found that within the stable periods there are often slowly growing or decaying trends. These trends provide additional information to a weighed value, when estimating parameters such as RTT and bandwidth. We therefore ran experiments using different methods to see how well they captured these trends. Our measurements show that a median or mean calculation wrongly estimates the value when there is a trend in the data set.
200
400
600
800
1000
1200
1400
1
3
5
7
9
11
13
15
17
19
21
23
25
27
29
31
33
35
37
39
41
43
45
47
49
51
Measurement
T i m e (m s )
Figure 8. RTT subplot
As an example, Figure 8 shows the region 144-197 extracted from the data in Figure 6. The figure also shows the mean, median and least squares linear estimation of the RTT value. It can be seen in the figure how the median and mean calculations underestimate the trend whereas the linear approximation better captures it. Until the Internet traffic has been mathematically modelled properly, it is impossible to determine the best method of estimation. We can however conclude that mean and median values are not suitable for estimating values such as RTT and bandwidth over a period since they do not accurately capture the trends.
6 Conclusion and future work
In this paper, we discussed how network monitoring should be carried out in a mixture of wired and wireless environments. The purpose of this discussion is to form guidelines for an overall framework for network monitoring taking into account the different needs of different applications. We have proposed a split network monitoring approach with passive/active monitoring. Central to this framework is the method of dynamic windowing to find a region with a relevant set of data on which to conduct statistical computations. We have also discussed the need for flexible filtering of the monitoring information in order to derive the different information needed by different applications. Finally, we have conducted initial experiments to verify our framework and to show the advantages of the dynamic window scheme.
Our experimental results show that using the dynamic windowing scheme gave consistently better results than using instantaneous measurements or fixed windows. Instantaneous measurements are not reliable when there is a large fluctuation in the measurements, and it is impossible to determine a correct size for the fixed window. We also believe that the dynamic windowing scheme is sufficient to determine the state of the network and to extract appropriate regions. We therefore conclude that the scheme is a universal tool that applies to all network types.
In the future, we intend to focus on statistical computations of monitoring results. Firstly, we will investigate possible ways of determining the optimal time window size for filtering transient events while still reacting on sustained bandwidth shifts. The combination of statistical properties and application requirements will be of primary interest. Secondly, we will work on statistical methods for use in different filters in order to provide accurate and for applications appropriate estimates of the performance parameters. Thirdly, we will work on implementation issues such as a protocol between a terminal and a SPAND server for specifying filter options etc. Finally, we intend to implement a prototype of the framework in order to evaluate its usefulness.
Acknowledgements
The authors would like to acknowledge Ericsson Radio Systems AB, Sweden and Ericsson Australia for their financial support.
References
[1]S. Seshan, M. Stemm and R. Katz. “SPAND:Shared Passive Network Performance
Discovery”. In Proc 1st Usenix Symposium on Internet Technologies and Systems
(USITS '97) Monterey, CA December 1997.
[2]https://www.doczj.com/doc/095899899.html,/xorg/nmtf/nmtf-tools.html.
[3]W. R. Stevens, TCP/IP Illustrated Volume 1 Addison-Wesley, Reading MA, Nov 1994.
[4]V. Paxon. “Measurements and Analysis of End-to-End Internet Dynamics. PhD thesis, U.
C. Berkeley, May 1996.
[5]R. L. Carter and M. E. Crovella. “Measuring bottleneck-link speed in packet switched
networks”. Technical Report BU-CS-96-007, Computer Science Department, Boston
University, March 1996.
[6]S. Keshav. “Packet-Pair Flow Control”. IEEE/ACM Transactions on Networking, Feb
1995.
[7]Pathchar – A Tool to Infer Characteristics in Internet Paths. ftp://https://www.doczj.com/doc/095899899.html,/pathchar,
1997.
[8]IEEE RFC 1470 June 1993
[9]M. Mathis and J. Mahdavi “Diagnosing Internet Congestion with a Transport Layer
Performance Tool”. In Proc. INET ’96, Montreal Canada, June 1996.
[10] D. Clark and D. Tannenhouse “Architectural considerations for a new generation of
protocols”. In SIGCOM’90 Philadelphia Pennsylvania September 1990.
[11]K. Lai and M. Baker, "Measuring Bandwidth." Proceedings of IEEE Infocom'99, March
1999.
[12]https://www.doczj.com/doc/095899899.html,/promotions/wireless
[13]https://www.doczj.com/doc/095899899.html,/products
[14]R. Katz and E. Brewer, “The Case for Wireless Overlaid Networks”, SPIE Multimedia
and Networking Conference (MMNC’96) San Jose January 1996.
[15]R. Braden et al. “Integrated Services in the Internet Architecture: an Overview”, RFC
1633, ISI, MIT and PARC, June 1994.
[16]K. Nichols, V. Jacobson and L. Zhang, “A two-bit Differential Services Architecture for
the Internet", IETF Internet draft November 1997.
一、对等网简介 “对等网”也称“工作组网”，那是因为它不像企业专业网络中那样是通过域来控制的，在对等网中没有“域”，只有“工作组”，这一点要首先清楚。正因如此，我们在后面的具体网络配置中，就没有域的配置，而需配置工作组。很显然，“工作组”的概念远没有“域”那么广，所以对等网所能随的用户数也是非常有限的。在对等网络中，计算机的数量通常不会超过20台，所以对等网络相对比较简单。在对等网络中，对等网上各台计算机的有相同的功能，无主从之分，网上任意节点计算机既可以作为网络服务器，为其它计算机提供资源；也可以作为工作站，以分享其它服务器的资源；任一台计算机均可同时兼作服务器和工作站，也可只作其中之一。同时，对等网除了共享文件之外，还可以共享打印机，对等网上的打印机可被网络上的任一节点使用，如同使用本地打印机一样方便。因为对等网不需要专门的服务器来做网络支持，也不需要其他组件来提高网络的性能，因而对等网络的价格相对要便宜很多。 对等网主要有如下特点： （1）网络用户较少，一般在20台计算机以内，适合人员少，应用网络较多的中小企业； （2）网络用户都处于同一区域中； （3）对于网络来说，网络安全不是最重要的问题。 它的主要优点有：网络成本低、网络配置和维护简单。 它的缺点也相当明显的，主要有：网络性能较低、数据保密性差、文件管理分散、计算机资源占用大。 二、对等网结构 虽然对等网结构比较简单，但根据具体的应用环境和需求，对等网也因其规模和传输介质类型的不同，其实现的方式也有多种，下面分别介绍： 1、两台机的对等网 这种对等网的组建方式比较多，在传输介质方面既可以采用双绞线，也可以使用同轴电缆，还可采用串、并行电缆。所需网络设备只需相应的网线或电缆和网卡，如果采用串、并行电缆还可省去网卡的投资，直接用串、并行电缆连接两台机即可，显然这是一种最廉价的对等网组建方式。这种方式中的“串／并行电缆”俗称“零调制解调器”，所以这种方式也称为“远程通信”领域。但这种采用串、并行电缆连接的网络的传输速率非常低，并且串、并行电缆制作比较麻烦，在网卡如此便宜的今天这种对等网连接方式比较少用。 2、三台机的对等网
物联网技术与应用 对等网络配置及网络资源共享 实验报告 组员：
1.实验目的 （1）了解对等网络基本配置中包含的协议，服务和基本参数 （2）了解所在系统网络组件的安装和卸载方法 （3）学习所在系统共享目录的设置和使用方法 （4）学习安装远程打印机的方法 2.实验环境 Window8，局域网 3.实验内容 （1）查看所在机器的主机名称和网络参数，了解网络基本配置中包含的协议，服务和基本参数 （2）网络组件的安装和卸载方法 （3）设置和停止共享目录 （4）安装网络打印机 4.实验步骤 首先建立局域网络，使网络内有两台电脑 （1）“我的电脑”→“属性”，查看主机名，得知两台计算机主机名为“idea-pc”和“迦尴专属”。 打开运行输入cmd，进入窗口输入ipconfig得到相关网络参数。局域网使用的是无线局域网。 （2）网络组件的安装和卸载方法：“网络和共享中心”→“本地连接”→“属
性”即可看到网络组件，可看其描述或卸载。 “控制面板”→“卸载程序”→“启用和关闭windows功能”，找到internet 信息服务，即可启用或关闭网络功能。 （3）设置和停止共享目录（由于windows版本升高，加强了安全措施和各种权
限，所以操作增加很多） 使用电脑“idea-pc”。“打开网络和共享中心”→“更改高级选项设置”。将专用网络，来宾或公用，所有网络中均选择启用文件夹共享选项，最下面的密码保护项选择关闭，以方便实验。 分享文件夹“第一小组实验八”，“右键文件夹属性”→“共享”→“共享”，选择四个中的一个并添加，此处选择everyone，即所有局域网内人均可以共享。
对等网络(P2P 一、概述 (一定义 对等网络(P2P网络是分布式系统和计算机网络相结合的产物,在应用领域和学术界获得了广泛的重视和成功,被称为“改变Internet的新一代网络技术”。 对等网络(P2P:Peer to Peer。peer指网络结点在: 1行为上是自由的—任意加入、退出,不受其它结点限制,匿名; 2功能上是平等的—不管实际能力的差异; 3连接上是互联的—直接/间接,任两结点可建立逻辑链接,对应物理网上的一条IP路径。 (二P2P网络的优势 1、充分利用网络带宽 P2P不通过服务器进行信息交换,无服务器瓶颈,无单点失效,充分利用网络带宽,如BT下载多个文件,可接近实际最大带宽,HTTP及FTP很少有这样的效果 2、提高网络工作效率 结构化P2P有严格拓扑结构,基于DHT,将网络结点、数据对象高效均匀地映射到覆盖网中,路由效率高 3、开发了每个网络结点的潜力 结点资源是指计算能力及存储容量,个人计算机并非永久联网,是临时性的动态结点,称为“网络边缘结点”。P2P使内容“位于中心”转变为“位于边缘”,计算模式由“服务器集中计算”转变为“分布式协同计算”。
4、具有高可扩展性(scalability 当网络结点总数增加时,可进行可扩展性衡量。P2P网络中,结点间分摊通信开销,无需增加设备,路由跳数增量小。 5、良好的容错性 主要体现在:冗余方法、周期性检测、结点自适应状态维护。 二、第一代混合式P2P网络 (一主要代表 混合式P2P网络,它是C/S和P2P两种模式的混合;有两个主要代表: 1、Napster——P2P网络的先驱 2、BitTorrent——分片优化的新一代混合式P2P网络 (二第一代P2P网络的特点 1、拓扑结构 1混合式(C/S+P2P 2星型拓扑结构,以服务器为核心 2、查询与路由 1用户向服务器发出查询请求,服务器返回文件索引 2用户根据索引与其它用户进行数据传输 3路由跳数为O(1,即常数跳 3、容错性:取决于服务器的故障概率(实际网络中,由于成本原因,可用性较低。
对等网络的网络弹性分析 摘要：网络弹性研究的是网络在节点失效或被有意攻击下所表现出来的特征。分析Gnutella网络的网络弹性，包括对于随机攻击的容错性和对于选择性攻击的抗攻击性，并与ER模型和EBA模型进行了对比。Gnutella网络对于随机攻击具有很好的容错性，但是对于选择性攻击却显得脆弱。最后对网络弹性进行了理论分析，给出了网络在出现最大集团临界点之前的平均集团大小的公式解。 关键词：对等网络；无标度；网络弹性；脆弱性 中图分类号：TP393.02文献标识码：A 文章编号：1001-9081(2007)04-0784-04 0 引言 在过去的40多年里，科学家习惯于将所有复杂网络看作是随机网络。随机网络中绝大部分节点的连结数目会大致相同。1998年开展的一个描绘互联网的项目却揭示了令人惊诧的事实：基本上，互联网是由少数高连结性的页面串联起来的，80%以上页面的连结数不到4个，而只占节点总数不到万分之一的极少数节点，例如门户网Yahoo和搜索引擎Google等类似网站，却高达上百万乃至几十亿个链接。研究者把包含这种重要集散节点的网络称为无标度网络[1]。
具有集散节点和集群结构的无标度网络，对意外故障具有极强的承受能力，但面对蓄意的攻击和破坏却不堪一击[2]。在随机网络中，如果大部分节点发生瘫痪，将不可避免地导致网络的分裂。无标度网络的模拟结果则展现了全然不同的情况，随意选择高达80%的节点使之失效，剩余的网络还可能组成一个完整的集群并保持任意两点间的连接，但是只要5%―10%的集散节点同时失效，就可导致互联网溃散成孤立无援的小群路由器。 许多复杂网络系统显示出惊人的容错特性，例如复杂通信网络也常常显示出很强的健壮性，一些关键单元的局部失效很少会导致全局信息传送的损失。但并不是所有的网络都具有这样的容错特性，只有那些异构连接的网络，即无标度网络才有这种特性，这样的网络包括WWW、因特网、社会网络等。虽然无标度网络具有很强的容错性，但是对于那些有意攻击，无标度网络却非常脆弱。容错性和抗攻击性是通信网络的基本属性，可以用这两种属性来概括网络弹性。 对等网络技术和复杂网络理论的进展促使对现有对等 网络的拓扑结构进行深入分析。对网络弹性的认识可以使从网络拓扑的角度了解网络的脆弱点，以及如何设计有效的策略保护、减小攻击带来的危害。本文研究Gnutella网络的网络弹性，并与ER模型和EBA模型进行了比较，对比不同类 型的复杂网络在攻击中的网络弹性。当网络受到攻击达到某
实验一建立对等网 一、实验目的 (1)熟悉10BASE-T星型拓扑以太网的网卡、线缆、连接器等网络硬件设备; (2)熟悉WINDOWS中的网络组件及各参数的设置; (3)理解对等网络的特点。 二、实验环境 此实验的基本要求就是两台以上计算机作为一个工作组,连接到一台服务器上,建立一个基于Windows的对等网络,物理结构为10BASE-T以太网。各工作组中的用户可以共享资源。 三、实验内容 (1)网络布线 EIA/TIA的布线标准中规定了两种双绞线的线序568A与568B,分别为: T568A:白绿 | 绿 | 白橙 | 蓝 | 白蓝 | 橙 | 白棕 | 棕 T568B:白橙 | 橙 | 白绿 | 蓝 | 白蓝 | 绿 | 白棕 | 棕 在整个网络布线中应用一种布线方式,但两端都有RJ45端头的网络连线无论就是采用端接方式A,还就是端接方式B,在网络中都就是通用的。实际应用中,大多数都使用T568B的标准,通常认为该标准对电磁干扰的屏蔽更好,本次实习中即采用了端接方式B。 (2)连接网线,建立对等网 连接网线的方式与网卡接口与网络结构有关,本局域网中采用的就是星型结构。以集线器(HUB)为中央结点,网络中所有计算机都通过双绞线连接至集线器,通过集线器交换信息。星型结构的优点就是利用中央结点可方便地提供服务与重新配置网络,单个连接点的故障只影响一个设备;缺点就是每个站点直接与中央结点相连,需要大量电缆,费用较高。 连接好网络线后接通计算机电源,观察网卡后面板上的两只LED工作状态指示灯。绿灯亮表示网络线接通,红灯间接闪烁说明网卡工作正常。 (3)MS-DOS方式中,执行ping命令进行测试
实验：Windows下对等网络的实现 实验目的 了解虚拟机中网络适配器的配置。 了解对等网络的基本概念与结构。 学习对等网络组件在系统中的安装与设置方法。 掌握利用对等网络进行数据传输与文件共享的方法。 实验任务 为虚拟机中的Windows系统安装网络组建。 配置虚拟机的IP地址设置、协议与服务绑定等网络参数，使之能从主机上访问虚拟机。 在虚拟机上设立共享目录、共享文件、设立权限等； 从主机映射虚拟机上的共享目录。 预备知识 虚拟机的原理和vmware虚拟机软件的使用 Windows操作系统基本操作 计算机网络原理与TCP/IP协议 实验平台与工具 PC机+Windows XP SP2 Vmware5.5+Windows2003 server 实验用时 120分钟 基础知识 VMware的几种网络模式 1.bridged(桥接模式) ：等于让Guest和Host系统并列在同一个子网中，占用两个ip，相互独立（对 于绑定网卡的子网络就不适用了，而且Guest的包就直接出去了，Host管不了）。等效网络结构 如下：
2.host-only(主机模式) ：对应Host里面的"VMnet1"，Guest的ip由VMware的DHCP提供，相当 于与Host网线直连如果要访问外网，还需要手工做网桥。其等效网络结构图如下： 3.NA T (网络地址转换模式) ：这个对于让Guest OS访问Internet是最简单的，对应"VMnet8"，直 接使用Guest认的网卡就行了，ip是VMware的DHCP对应VMnet8分配的，与外网无关，但 Guest对外的访问，会自动转换出去。其等效网络结构图如下： 实验步骤 第一部分：准备运行环境
第19卷第5期湖 北 工 学 院 学 报2004年10月 Vol.19No.5 Journal of Hubei Polytechnic University Oct.2004 [收稿日期]2004-05-25 [作者简介]詹春华(1971-),男,湖北红安人,华中科技大学硕士研究生,研究方向:计算机网络,分布式计算. [文章编号]1003-4684(2004)10 0034 03 对等网络搜索方法比较与分析 詹春华,陈晓苏 (华中科技大学计算机学院,湖北武汉430074) [摘 要]详细介绍了现存的P2P 网络中的搜索技术,对搜索方法进行了比较和分析,指出了它们的优缺点.[关键词]对等网络;分布式搜索;搜索[中图分类号]T P 393 [文献标识码]:A 对等网络(peer to peer,P2P)技术是近年来计算机网络技术中的一个热点.P2P 可简单地定义为对等点(peer)之间通过直接交换信息从而达到共享计算机资源和服务,每一个对等点可以同时充当客户端和服务器两种角色,以该技术构建的网络称为对等网.对等网络是一个完全分布式的网络,所有对等点都是自治的,没有统一的管理,它们共同组成一个系统.对等网络在容错性、资源共享的可扩展性、自我组织、负载平衡、匿名等方面具有很大的优势.目前P2P 技术被广泛应用于文件共享、协同工作、分布式计算等领域. 对等网络中的一个基本问题就是如何找到储存有特定数据的节点,即分布式搜索问题.当节点在其自身找不到想要的对象时,就会发出请求,搜索过程涉及请求转发方法、收到请求消息的节点、消息的形式、某些节点维护的局部索引等方面. 不同网络结构可能会采用不同搜索方法.搜索方法对于对等网络系统的性能、网络流量和可扩展性等方面有很大影响. 笔者详细介绍了现存的P2P 网络中的搜索技术,对搜索方法进行了比较和分析,指出了其优缺点. 1 P 2P 系统分类 当前的P2P 系统,可以根据系统是否对拓扑结构和共享信息(文件)存放位置作出规定分为两大类:非结构化系统和高度结构化系统. 非结构化系统对网络拓扑结构的构成没有严格 的限制,节点可以自由地动态加入网络.非结构化系统主要关注的是共享数据,但对每个节点的共享数据的存放位置没有很好的规则,每个节点可以随意地决定其要共享的数据和共享数据的位置.非结构化系统不能保证每个搜索都能成功. 非结构化系统还可根据P2P 网络模型分为两类:纯P2P 系统和混合P2P 系统.在纯P2P 系统中每个节点的地位都是平等的.混合P2P 系统中,某些节点为超级节点,其余节点则为叶节点,超级节点为其相邻的叶节点的文档建立索引,并为相邻的叶节点提供搜索服务.这一类的系统有Napster [1]、Gnutella [2] 、Fr eenet [3] 等. 高度结构化系统对拓扑结构的 叠加"被严格控制,文件(或者文件指针)存放在确定的位置上. 2 非结构化系统的搜索方法 非结构化系统的搜索方法主要有两类:一类为盲目搜索,它不依赖于任何已知信息,简单地将搜索请求传播给足够多的节点.另一类为启发式搜索,节点利用已知的信息进行搜索.已知信息可以是节点根据已有的搜索结果逐步建立的搜索知识库,也可能是准确的目标位置信息.这些信息的位置也有很大变化,在集中式网络结构中,该信息存在于一个所有节点都知道的中央目录,在分布式网络结构中,该信息保存在每个节点自身. 2.1 盲目搜索 2.1.1 基本盲目搜索方法 这种方法基于洪泛法,节点向所有相邻节点转发搜索请求,在搜索请求中
实验小型办公室对等网络的组建 一．实验目的 1．熟悉网卡、掌握如何在Windows下如何察看网卡的型号、MAC地址、IP 地址等参数。 2．熟悉Windows中的网络组建及各参数的设置和基本意义。 3．在对等网中建立共享。 4．网络测试命令PING的用法 二．实验任务与要求 1、网卡是网络中不可缺少的网络设备，掌握其使用情况，及如何设置其参数对网络的正常使用非常重要。本部分要完成以下任务： （1）利用Windows下Ipconfig 命令查看网卡的基本参数。 （2）如何设置网卡的IP地址。 2、Ping是个使用频率极高的实用程序，用于确定本地主机是否能与另一台主机交换（发送与接收）数据报。根据返回的信息，就可以推断TCP/IP参数是否设置得正确以及运行是否正常。 （1）Ping 本机Ip（Ping 本机机器名；Ping 127.0.0.1）。 （2）Ping 邻近计算机的ip（或者是对方计算机的机器名）。 （3）Ping 网站（前题是能接入Internet）。 3、对等网络（Peer to Peer）也称工作组模式，其特点是对等性，即网络中计算机功能相似，地位相同，无专用服务器，每台计算机相对网络中其他的计算机而言，既是服务器又是客户机，相互共享文件资源以及其他网络资源。本次实验要求完成以下任务： （1）如何修改计算机所在工作组、计算机名。 （2）配置网卡并注意观察网络硬件的连接方法。 （3）完成对等网的组建与测试。 三．实验步骤 （1）Ipconfig的使用
（2）利用“网上邻居”修改网络参数。 （3）PING命令的使用（报告中只要求简化过程） ·ping 本机IP
对等网络概念原理及应用 一、对等网络概述 1.1 对等网络的定义 集中式管理与控制模式网络结点在功能上是不平等的，总有一些少数结点在网络中占居中心主导地位，管理其它从属结点可执行的操作，控制它们间的信息交互。其特点是网络管理复杂度小且易于控制，但不足是网络工作效率低、规模扩展性不好且存在单点故障。 分布式管理与控制模式网络结点在功能上是平等的，网络中没有结点比其它结点拥有更大的特权，网络的管理与控制有各个结点间的相互协作来完成的。其特点是网络工作效率高、规模扩展性强，但网络管理复杂度大且不易于控制。 虽着因特网规模的不断扩大与应用功能的不断扩展，目前因特网主流应用模式---客户服务器模式已不能满足用户的实际需求。迫切需要新的应用模式。 对等计算思想本质：打破传统的网络集中管理与控制的客户服务器模式，使网络成员享有自由、平等、互联的权力。 对等网络（Peer-to-Peer network，P2P）：分布式系统与计算机网络相结合的产物，是采用对等计算模式工作的计算机网络（应用服务）。每个节点在行为上是自由的，在功能上是平等的，在关系上是互联的。所有节点分布式地自组织成一个整体网络，能够极大地提高网络效率，充分地利用网络带宽，发挥每个网络节点的潜能。 对等计算起源于1956年，但由于当时的网络技术条件的限制与网络应用没有充分地发展起来，一直未引起学术、工业与商业界重视，因此，它的发展十分缓慢。 网络应用实践表明：任何一种思想、理论与技术的流行通常需要一个杀手级的应用（killer application）,以一种征服性的力量冲击人们的传统思维。对等网络的杀手锏级的应用则出现在1999年，第一个应用性对等网络Napster，一种音乐下载软件，创造了在半年时间内拥有5000万用户的网络奇迹，向世人展示了对等网络的优异性能和巨大潜力。继Napster之后，又有一系列随着人们耳熟能详的对等网络应用软件，如Gnutella、KaZaA、BitTorrent、eDonkey/eMule、Skype