电气工程及自动化文献综述 (21).docx

《电气工程及自动化文献综述 (21).docx》由会员分享，可在线阅读，更多相关《电气工程及自动化文献综述 (21).docx（19页珍藏版）》请在淘文阁 - 分享文档赚钱的网站上搜索。

1、中国地质大学长城学院本科毕业论文外文资料翻译院 别工程技术系学生姓名连昊阳专 业电气工程及其自动化学 号0431302082017年3月 20 日网络多播（组播）概述H3C技术论坛1概述1.1产生背景传统的IP通信有两种方式：一种是在源主机与目的主机之间点对点的通信，即单播；另一种是在源主机与同一网段中所有其它主机之间点对多点的通信，即广播。如果要将信息发送给多个主机而非所有主机，若采用广播方式实现，不仅会将信息发送给不需要的主机而浪费带宽，也不能实现跨网段发送；若采用单播方式实现，重复的IP包不仅会占用大量带宽，也会增加源主机的负载。所以，传统的单播和广播通信方式不能有效地解决单点发送、多点

2、接收的问题。组播是指在IP网络中将数据包以尽力传送的形式发送到某个确定的节点集合（即组播组），其基本思想是：源主机（即组播源）只发送一份数据，其目的地址为组播组地址；组播组中的所有接收者都可收到同样的数据拷贝，并且只有组播组内的主机可以接收该数据，而其它主机则不能收到。1.2技术优点组播技术有效地解决了单点发送、多点接收的问题，实现了IP网络中点到多点的高效数据传送，能够大量节约网络带宽、降低网络负载。作为一种与单播和广播并列的通信方式，组播的意义不仅在于此。更重要的是，可以利用网络的组播特性方便地提供一些新的增值业务，包括在线直播、网络电视、远程教育、远程医疗、网络电台、实时视频会议等互联网

3、的信息服务领域。2组播技术实现组播技术的实现需要解决以下几方面问题：组播源向一组确定的接收者发送信息，而如何来标识这组确定的接收者？这需要用到组播地址机制；接收者通过加入组播组来实现对组播信息的接收，而接收者是如何动态地加入或离开组播组的？即如何进行组成员关系管理；组播报文在网络中是如何被转发并最终到达接收者的？即组播报文转发的过程；组播报文的转发路径（即组播转发树）是如何构建的？这是由各组播路由协议来完成的。2.1组播地址机制2.1.1IP组播地址IP组播地址用于标识一个IP组播组。IANA把D类地址空间分配给组播使用，范围从224.0.0.0到239.255.255.255。图1IP组播地

4、址格式如图1所示，IP组播地址前四位均为“1110”，而整个IP组播地址空间的划分则如图2所示。图2IP组播地址划分224.0.0.0到224.0.0.255被IANA预留，地址224.0.0.0保留不做分配，其它地址供路由协议及拓扑查找和维护协议使用。该范围内的地址属于局部范畴，不论TTL为多少，都不会被路由器转发；224.0.1.0到238.255.255.255为用户可用的组播地址，在全网范围内有效。其中232.0.0.0/8为SSM组地址，而其余则属于ASM组地址。有关ASM和SSM的详细介绍，请参见“2.5组播模型分类”一节；239.0.0.0到239.255.255.255为本地管

5、理组播地址，仅在特定的本地范围内有效，也属于ASM组地址。使用本地管理组地址可以灵活定义组播域的范围，以实现不同组播域之间的地址隔离，从而有助于在不同组播域内重复使用相同组播地址而不会引起冲突。说明：224.0.1.0/24网段内的一些组播地址也被IANA预留给了某些组播应用。譬如，224.0.1.1被预留给NTP（Network Time Protocol，网络时间协议）所使用。2.1.2IP组播地址到链路层的映射说明：本文只讨论以太网链路层协议的组播实现，其它链路层协议的组播实现并不作为本文讨论的重点。IANA将MAC地址范围01:00:5E:00:00:0001:00:5E:7F:FF:

6、FF分配给组播使用，这就要求将28位的IP组播地址空间映射到23位的组播MAC地址空间中，具体的映射方法是将组播地址中的低23位放入MAC地址的低23位，如图3所示。图3IP组播地址到组播MAC地址的映射由于IP组播地址的后28位中只有23位被映射到组播MAC地址，这样会有32个IP组播地址映射到同一组播MAC地址上。2.2组成员关系管理组成员关系管理是指在路由器/交换机上建立直联网段内的组成员关系信息，具体说，就是各接口/端口下有哪些组播组的成员。2.2.1IGMPIGMP运行于主机和与主机直连的路由器之间，其实现的功能是双向的：一方面，主机通过IGMP通知路由器希望接收某个特定组播组的信息

7、；另一方面，路由器通过IGMP周期性地查询局域网内的组播组成员是否处于活动状态，实现所连网段组成员关系的收集与维护。通过IGMP，在路由器中记录的信息是某个组播组是否在本地有组成员，而不是组播组与主机之间的对应关系。目前IGMP有以下三个版本：IGMPv1（RFC 1112）中定义了基本的组成员查询和报告过程；IGMPv2（RFC 2236）在IGMPv1的基础上添加了组成员快速离开的机制等；IGMPv3（RFC 3376）中增加的主要功能是成员可以指定接收或拒绝来自某些组播源的报文，以实现对SSM模型的支持。以下着重介绍IGMPv2的原理。图4IGMPv2的工作原理如图4所示，当同一个网段内

8、有多个IGMP路由器时，IGMPv2通过查询器选举机制从中选举出唯一的查询器。查询器周期性地发送普遍组查询消息进行成员关系查询，主机通过发送报告消息来响应查询。而作为组成员的路由器，其行为也与普通主机一样，响应其它路由器的查询。当主机要加入组播组时，不必等待查询消息，而是主动发送报告消息；当主机要离开组播组时，也会主动发送离开组消息，查询器收到离开组消息后，会发送特定组查询消息来确定该组的所有组成员是否都已离开。通过上述机制，在路由器里建立起一张表，其中记录了路由器各接口所对应子网上都有哪些组的成员。当路由器收到发往组G的组播数据后，只向那些有G的成员的接口转发该数据。至于组播数据在路由器之间

9、如何转发则由组播路由协议决定，而不是IGMP的功能。2.3组播报文转发2.3.1组播转发树组播报文在网络中沿着树型转发路径进行转发，该路径称为组播转发树。它可分为源树（Source Tree）和共享树（RPT）两大类：1.源树源树是指以组播源作为树根，将组播源到每一个接收者的最短路径结合起来构成的转发树。由于源树使用的是从组播源到接收者的最短路径，因此也称为最短路径树（SPT）。对于某个组，网络要为任何一个向该组发送报文的组播源建立一棵树。源树的优点是能构造组播源和接收者之间的最短路径，使端到端的延迟达到最小。但付出的代价是，在路由器中必须为每个组播源保存路由信息，这样会占用大量的系统资源，路

10、由表的规模也比较大。2.共享树以某个路由器作为路由树的树根，该路由器称为汇集点（RP），共享树就是由RP到所有接收者的最短路路径所共同构成的转发树。使用共享树时，对应某个组网络中只有一棵树。所有的组播源和接收者都使用这棵树来收发报文，组播源先向树根发送数据报文，之后报文又向下转发到达所有的接收者。共享树的最大优点是路由器中保留的路由信息可以很少，缺点是组播源发出的报文要先经过RP，再到达接收者，经由的路径通常并非最短，而且对RP的可靠性和处理能力要求很高。2.4组播路由协议与单播路由一样，组播路由协议也分为域内和域间两大类：域内组播路由协议：根据IGMP协议维护的组成员关系信息，运用一定的组播

11、路由算法构造组播分发树，在路由器中建立组播路由状态，路由器根据这些状态进行组播数据包转发；域间组播路由协议：根据网络中配置的域间组播路由策略，在各自治系统间发布具有组播能力的路由信息以及组播源信息，使组播数据能在域间进行转发。2.4.1域内组播路由协议在众多域内组播路由协议中，PIM是目前较为典型的一个。按照转发机制的不同，PIM可以分为DM（Dense Mode，密集模式）和SM（Sparse Mode，稀疏模式）两种模式。1.PIM-DM在PIM-DM域中，运行PIM-DM的路由器周期性地发送PIM Hello消息，以发现邻接的PIM路由器，进行叶子网络、叶子路由器的判断，并负责在多路访问

12、网络中选举DR尽管PIM-DM本身并不需要DR，但如果PIM-DM域中的共享网络上运行了IGMPv1，则需要选举出DR来充当共享网络上的IGMPv1查询器。PIM-DM属于密集模式的组播路由协议，使用“推”模式传送组播数据，通常适用于组播组成员相对比较密集的小型网络，其基本原理如下：PIM-DM假设网络中的每个子网都存在至少一个组播组成员，因此组播数据将被扩散到网络中的所有节点。然后，PIM-DM对没有组播数据转发的分支进行剪枝，只保留包含接收者的分支。这种“扩散剪枝”现象周期性地发生，被剪枝的分支也可以周期性地恢复成转发状态。当被剪枝分支的节点上出现了组播组的成员时，该节点通过主动向其上游发

13、送嫁接报文，从而由剪枝状态恢复成转发状态，以恢复对组播数据的转发。2.PIM-SM在PIM-SM域中，运行PIM-SM的路由器周期性地发送PIM Hello消息，以发现邻接的PIM路由器，并负责在多路访问网络中选举DR。这里，DR负责为与其直连的组成员向组播树根节点的方向发送加入/剪枝消息，或是将直连组播源的数据发向组播分发树。PIM-SM属于稀疏模式的组播路由协议，使用“拉”模式传送组播数据，通常适用于组播组成员分布相对分散、范围较广的大中型网络，其基本原理如下：PIM-SM假设所有主机都不需要接收组播数据，只向明确提出需要组播数据的主机转发。PIM-SM实现组播转发的核心任务就是构造并维护

14、RPT，RPT选择PIM域中某台路由器作为公用的根节点RP，组播数据通过RP沿着RPT转发给接收者；连接接收者的路由器向某组播组对应的RP发送加入报文，该报文被逐跳送达RP，所经过的路径就形成了RPT的分支；组播源如果要向某组播组发送组播数据，首先由与组播源侧DR负责向RP进行注册，把注册报文通过单播方式发送给RP，该报文到达RP后触发建立SPT。之后组播源把组播数据沿着SPT发向RP，当组播数据到达RP后，被复制并沿着RPT发送给接收者。2.4.2域间组播路由协议域间组播路由用来实现组播信息在AS之间的传递，目前比较成型的解决方案有：MBGP：用于在自治域之间交换组播路由信息；MSDP：用于

15、在ISP之间交换组播源信息。1.MBGP域间路由的首要问题是路由信息（或者说可达信息）如何在自治系统之间传递，由于不同的AS可能属于不同的运营商，因此除了距离信息外，域间路由信息必须包含运营商的策略，这是与域内路由信息的不同之处。组播的网络拓扑和单播拓扑有可能不同，这里既有物理方面的原因，也有策略方面的原因。网络中的一些路由器可能只支持单播不支持组播，也可能按照策略配置不转发组播报文。为了构造域间组播路由树，除了要知道单播路由信息外，还要知道网络中哪些部分是支持组播的，即组播的网络拓扑情况。简而言之，域间的组播路由信息交换协议应该满足下面的要求：1.能对单播和组播拓扑进行区分；2.有一套稳定的

16、对等和策略控制方法。目前使用最多的域间单播路由协议是BGP-4，由于BGP-4已满足后一个条件，而且已被证明是一个有效的、稳定的单播域间路由协议，因此为了实现域间组播路由信息的传递，合理的解决方案就是对BGP-4协议进行增强和扩展，而不是构建一套全新的协议。在RFC 2858中规定了对BGP进行多协议扩展的方法，扩展后的BGP协议（即MP-BGP，也写作BGP-4+）不仅能携带IPv4单播路由信息，也能携带其它网络层协议（如组播、IPv6等）的路由信息，携带组播路由信息只是其中一个扩展功能，称为组播BGP（MBGP）。有了MBGP之后，单播和组播路由信息可以通过同一个进程交换，但是存放在不同的

17、路由表里。由于MBGP是BGP-4协议的一个增强版，因此BGP-4所支持的常见的策略和配置方法都可以用到组播里。2.MSDP在基本的PIM-SM模式下，组播源只向本PIM-SM域内的RP注册，且各域的组播源信息是相互隔离的，因此RP仅知道本域内的组播源信息，只能在本域内建立组播分发树，将本域内组播源发出的组播数据分发给本地用户。而对于ISP来说，不希望依靠其它ISP的RP转发组播流量，但同时又要求无论组播源的RP在哪里，都能从组播源获取信息发给自己内部的成员。MSDP就是为了解决多个PIM-SM域之间的互连而开发的一种域间组播解决方案，用来发现其它PIM-SM域内的组播源信息。MSDP通过将某

18、个域内的RP与其它域内的RP建立MSDP对等体关系，以连通各PIM-SM域的RP，并利用这些对等体关系交换组播源信息。尽管MSDP是为域间组播开发的，但它在PIM-SM域内还有着一项特殊的应用Anycast RP（任播RP）。Anycast RP是指在同一PIM-SM域内通过设置两个或多个具有相同地址的RP，并在这些RP之间建立MSDP对等体关系，以实现域内各RP之间的负载分担和冗余备份。图5PIM-SM单域组播组网3总结和展望组播技术从1988年提出至今已经历了20年的发展，许多国际组织对组播的技术研究和业务开展进行了大量的工作。在IP网络中多媒体业务日渐增多的情况下，组播技术为多媒体业务的

19、开展提供了传输基础。组播技术涵盖了从地址方案、成员管理和路由建立等各个方面，其中组播地址的分配方式、域间组播路由以及组播安全等仍是研究的热点。从目前的情况看，组成员管理技术普遍采用IGMPv2；PIM-SM因其良好的扩展性以及从RPT向SPT切换的能力而成为域内组播路由技术的首选；域间组播路由协议现阶段普遍采用PIM-SM/MBGP/MSDP的组合方案。组播技术可以提供包括流媒体、视频会议、IPTV等在内的各种宽带增值业务，但这些业务的顺利开展还依赖于有效的业务管理、监控及安全控制。结合在业务运营管理方面的理解和经验积累，H3C公司提供不断完善的可运营、可管理的组播解决方案，我们将继续致力于推

20、动组播技术的发展、组播业务的普及和功能的完善。Overview of network multicast (multicast)H3C Technology Forum1 Overview1.1 backgroundThe traditional IP communication in two ways: one is communication, point-to-point between the source host and the destination host is unicast; the other is on the source host and the same seg

21、ment between all other hosts point to multipoint communication, namely broadcast. If you want to send information to a plurality of host and non host all, if the broadcast mode, not only will send the information to the host does not need to waste bandwidth, also can cross segment transmission; if t

22、he unicast, repeat IP packet not only will occupy a lot of bandwidth, load will increase the source host the. Therefore, the traditional unicast and broadcast communication can not effectively solve the problem of single point sending and receiving.Refers to the multicast network IP data packet to s

23、end to send the form to a certain set of nodes (multicast group), the basic idea is: the source host (i.e. multicast source) only send a copy of the data, the destination address is the multicast group address; all receivers in the multicast group can be received the same copy of the data, and only

24、the host in the multicast group can receive the data, while the other host is not received.1.2 technical advantagesMulticast technology can effectively solve the problem of single point sending and receiving, and realize the high efficiency data transmission in IP network, which can save network ban

25、dwidth and reduce network load. As a way of communicating with unicast and broadcast, multicast is not only in the sense of. More importantly, the characteristics of multicast network can be used conveniently to provide some new value-added services, including online broadcast, network television, r

26、emote education, remote medical treatment, network radio, real-time video conferencing and Internet information services.2.The realization of the multicast technologyMulticast technology needs to solve the following problems:Send the recipient information l multicast source to a group of determined,

27、 and how to identify the set of receivers? - this requires a multicast address mechanism;by receivers join the multicast group to receive the multicast information, and the receiver is how to dynamically join or leave a multicast group? - how to manage group membership; multicast packet is to be for

28、warded and eventually arrive at the receiver in the network? Multicast packet forwarding process;The forwarding path of L multicast message (i.e. multicast forwarding tree) is how to construct? this is done by multicast routing protocols.2.1 multicast address system2.1.1 IP multicast addressIP multi

29、cast address is used to identify a IP multicast group. IANA assigns the D class address space to multicast, ranging from 224.0.0.0 to 239.255.255.255.Figure 1 IP multicast address formatAs shown in Figure 1, IP multicast address is the first four 1110, and the entire IP multicast address space is di

30、vided as shown in figure 2.Figure 2 IP multicast address division224.0.0.0 to 224.0.0.255 IANA 224.0.0.0 do not keep reservation, address allocation, address for other routing protocols and topology search and maintenance protocol. The scope of the address belongs to the local category, no matter ho

31、w much TTL, will not be forwarded by the router; 224.0.1.0 to 238.255.255.255 for the multicast address available to users, effectively within the whole network. Which 232.0.0.0/8 address for the SSM group, while the rest belong to the ASM group address. Details of ASM and SSM, please refer to the 2

32、.5 multicast model classification section; 239.0.0.0 to 239.255.255.255 for the local management of multicast address is valid only in specific local context, also belong to the ASM group. The use of local management group address can flexibly define multicast domain address range, in order to achie

33、ve isolation between different multicast domain, which contributes to the repeated use of the same multicast address in different multicast domain without conflict.Description &:Some multicast addresses in 224.0.1.0/24 segment are reserved by IANA for some multicast applications. For example, 224.0.

34、1.1 is reserved for use by NTP (Network Time Protocol, network time protocol).2.1.2 IP multicast address mapping to the link layerDescription :In this paper, we discuss the multicast implementation of Ethernet link layer protocol, and the multicast implementation of other link layer protocols is not

35、 the focus of this paper.IANA MAC address ranges from 01:00:5E:00:00:00 to 01:00:5E:7F:FF:FF allocated to the multicast, which requires the IP multicast address space mapping 28 bit to 23 bit address space of MAC multicast, mapping method in particular the multicast address in the low 23 bits into t

36、he MAC address of the lower 23, as shown in figure 3.Since only 23 bits of the last 28 bits of the IP multicast address are mapped to the multicast MAC address, the IP multicast address is mapped to the same multicast MAC address.2.2 group membership managementGroup membership management is to set u

37、p the group member relationship information in the direct link segment on the router / switch.2.2.1 IGMPIGMP is running on the host and the host directly connected to the router, its function is bidirectional: on the one hand, the host through the IGMP notification router want to receive a particula

38、r multicast group information; on the other hand, through the IGMP router periodically query the multicast group members within the LAN is active, the networking section the collection and maintenance of group membership. Through IGMP, the information recorded in the router is whether a multicast gr

39、oup. IGMPv1 (RFC 1112) in the definition of group membership query and report the basic process of; IGMPv2 (RFC 2236) added group members quickly left the mechanism on the basis of IGMPv1; IGMPv3 (RFC 3376) the main function is the increase in member can be specified to accept or reject messages fro

40、m some of the multicast source, to support the implementation of the SSM model.The following focuses on the principle of IGMPv2.As shown in Figure 4, when there are multiple IGMP routers in the same network segment, IGMPv2 selects the unique query from the query voting mechanism. The query periodica

41、lly sends a general set of query messages for membership queries, and the host responds to queries by sending a report message. As a member of the router, its behavior is the same as the general host, in response to other routers query.When the host to join the multicast group, without waiting for t

42、he query message, but take the initiative to send a report message; when the host will leave a multicast group, it will take the initiative to send a leave group message, inquiries received leave group message, sends a specific query message to determine all the group members of the group have been

43、left.Through the above mechanism, a table is set up in the router, which records the members of each group on the subnet corresponding to each interface of the router. When the router receives multicast data sent to group G, the data is forwarded only to the interface of the G member. As to how mult

44、icast data is transmitted between routers, it is decided by multicast routing protocol, not IGMP.2.3 multicast packet forwarding2.3.1 multicast forwarding treeMulticast packets are forwarded along the tree based forwarding paths in the network. It can be divided into two categories: source tree (Sou

45、rce Tree) and shared tree (RPT):1 source treeThe source tree is a forwarding tree composed of a multicast source as the root of the tree, which combines the shortest path from the multicast source to each receiver. Because the source tree uses the shortest path from the multicast source to the recei

46、ver, it is also called the shortest path tree (SPT). For a group, the network will set up a tree for any multicast source that sends messages to the group.The advantage of the source tree is that it can construct the shortest path between the multicast source and the receiver, so as to minimize the

47、end-to-end delay. But the price is that the router must save routing information for each multicast source, which will take up a large amount of system resources, the scale of the routing table is relatively large.2 shared treeA router is the root of the routing tree, which is called the rendezvous

48、point (RP), which is composed of RP and the shortest path of all receivers. When you use a shared tree, there is only one tree in a network. All of the multicast source and receiver use this tree to send and receive packets, the multicast source sends data packets to the root, then the message is forwarded to all receivers.The biggest advantage of share