【技术课件】多媒体技术

单击此处编辑母版标题样式,单击此处编辑母版文本样式,第二级,第三级,第四级,第五级,*,多媒体技术,1,第十章 网络组播技术,2,1,组播的概念,广播,：将源端数据发送到一个网络中的所有主机的传播方式。广播使用广播地址。,单播,：将源端数据发送到网络中的某个特定主机的传播方式。单播使用特定主机地址。,组播,：将源端数据发送到网络中的一组主机的传播方式。组播使用组地址。,3,1,组播的概念,Src,Src,组播：一种有效节省网络带宽的传输方式。,4,2,组播的应用,广泛应用于一对多（,one-to-many,）的传送需要,举例,:,新闻,/,体育消息,/,股票信息,/,天气信息的更新发布,远程教育（,Distance learning,）,网络路由信息更新（,routing updates,）,远程会议（,Teleconferencing,，,audio, video, shared whiteboard, text editor,）,分布式交互式游戏和仿真（,Distributed interactive gaming or simulations,）,内容发布（,Content distribution,）,;,软件发布（,Software distribution,）,Web,缓存更新（,Web-cache updates,）,数据库应用（,Database replication,）,5,3,组播与广播、单播的比较,广播（,Broadcast,）,:,将一个数据拷贝发送到网络中的所有主机,实现简单（,Simple,）,但是效率不高（,inefficient,）,即使主机对该数据不感兴趣也必须处理它,因此占用了主机的,CPU,资源,会产生广播风暴“,broadcast storms,”,多个重复的单播（,Replicated Unicast,）,发送者依次向各接收主机发送相同的数据,发送者必须事先知道每个接收主机的地址,通讯量在发送方过于集中，,但比较可靠（,Reliability,= per-receiver state, separate sessions/processes at sender,）,6,4 IP,组播的体系结构,Hosts,Routers,服务模型,Service model,主机与路由器关系协议,(IGMP),组播路由协议,Multicast routing protocols,7,5 IP,组播模型,: RFC 1112,发送者（源）不需要知道接收主机,使用组地址“,group address,”,标识一个群组,即,IP,包的头部使用的是组地址而不是单播地址,群组所包含的主机数量是任意的,（,any size,）,;,群组成员可以位于互联网任何地方,群组关系（,Group membership,）是非显式知道的（,not explicitly known,）,接收主机可以在任何时候加入或者退出群组通信,（,Receivers,can,join/leave at will,）,8,网络负责构造分配树（,Distribution tree,，也叫组播树 ）以实现组播传输,在组播传输中，任何子网中不能出现同一数据的两个副本,组播数据只能传送给群组成员（由于群组成员可以在任何时候加入或者退出群组，因此组播传送是一个非常动态化的问题（,multicast delivery tree changes dynamically,）,5 IP,组播模型,: RFC 1112,9,6 IP,组播地址,IP,组播使用,D,类,IP,地址,（,Class D IP addresses,）,224.0.0.0 239.255.255.255,组地址的分配（,Address allocation,）,:,IANA,指定了一些永久地址,（,Well-known : 224.0.0.x and 224.0.1.x,）；,其他组地址是临时组地址,（,Transient multicast addresses,），,需要动态分配,每一个组地址代表一个主机群组,（,host group,）,IP,组地址是非结构化的,（,flat address space,）,1,1,1,0,Group ID,10,IP,组播服务：组播数据的接收,提供两种操作,加入组播群组,:,Join-IP-Multicast-Group(group-address, interface),退出群组：,Leave-IP-Multicast-Group(group-address, interface),加入群组后，主机通过常规,IP,接收操作（,normal IP-receive operation,）接收组播数据包,11,7 IP,组播地址到链路层组地址的映射,以太网（,Ethernet,）等局域网使用,802,地址,:,采用直接映射,（,Direct mapping!,较,unicast,简单,!,无需,ARP,等协议,.,）,LAN multicast address,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,1,0,1,1,1,1,0,0,1,1,1,0,28 bits,23 bits,IP multicast address,Group bit,12,IP Multicast Architecture,Hosts,Routers,Service model,Host-to-router protocol(IGMP),Multicast routing protocols(various),13,8,群组关系管理协议（,Internet Group Management Protocol,）,IGMP:,在子网中建立、维护和撤销群组的信号协议,“,signaling protocol,”,。,目标,:,保持路由器对整个局域网上群组关系的更新,（,keep router up-to-date with group membership of entire LAN,）,但是，路由器不需要知道所有成员的存在，而只需要知道是否有成员主机存在,14,IGMP,协议的工作原理,一个网络选择一个路由器充当查询者,“querier”,查询者周期性发送群组关系查询报文,IGMP,（,Membership Query message,）,到一个特定地址,（,all-systems group,，,224.0.0.1),所有主机接收到该报文后，启动一个随机计时器,（,random timers,，,010,秒,),Q,Routers:,Hosts:,15,IGMP,协议的工作原理,(cont.),一旦有一个主机超时，它就发送一个群组关系应答报文,（,Membership Report,）,到组,G,其他主机也接收到该报文，立即终止其定时器,路由器负责监听所有群组的应答报文，如果一个群组未给出应答（超时），路由器就终止对该群组报文的转发,Q,G,G,G,G,Routers:,Hosts:,16,IP Multicast Architecture,Hosts,Routers,Service model,Host-to-router protocol(IGMP),Multicast routing protocols,17,9,组播路由（,Multicast Routing,）,组播路由的基本目标：在网络层为传送组播数据包而建立组播分配树,组播树的叶结点是指包含有群组成员主机的子网,(detected by IGMP),组播路由是一个比较困难的课题,（,Multicast service model makes it hard,）,组播的匿名性,动态加入和退出（,Dynamic join/leave,）,18,简单组播路由技术概览,基于泛播与剪枝技术的路由,首先在整个网络泛播组播数据包,然后剪除不含成员主机的网络分枝,属于这类技术的协议,:,DVMRP, PIM-DM,基于链路状态的组播路由协议,一个路由器知道有成员主机加入群组时，它就在整个网络广播该群组关系,一个路由器接收到组播数据时，它就利用,Dijkstra,算法计算到目标主机的最短路经，并转发组播数据,协议举例,:,MOSPF,19,基于泛播的组播路由过程,G,G,S,G,20,泛播（,Flooding,）,G,G,S,G,21,剪枝（,Pruning,）,G,G,S,Prune (s,g),Prune (s,g),G,22,Graft (s,g),Graft (s,g),嫁接（,Grafting,）,G,G,S,G,G,Report (g),23,嫁接完成后,G,G,S,G,G,24,距离向量组播路由（,Distance-Vector Multicast Routing,）,DVMRP,由两部分组成,:,常规距离向量协议,(like RIP),组播数据包转发协议,DVMRP,转发组播数据包,if,The packet,arrived from the link used to reach the source of the packet,即反向路径转发, RPF,仅在子链路上转发组播数据包,如果下游链路没有成员节点，则进行剪枝处理,25,DVMRP,协议的缺点,像其他基于距离向量的协议一样,受无穷计数（,count-to-infinity,）和环路影响（,transient looping,）,伸缩性受到类,RIP,协议的制约,.,也引入了新的伸缩性问题（,scaling limitations,）,:,路由器的,(S,G),状态，使路由器存储开销比较大,!,广播对网络冲击比较大,.,26,Multicast Backbone (MBone),An,overlay network,of IP multicast-capable routers using DVMRP,Tools: sdr (session directory), vic, vat, wb,Host/router,MBone router,Physical link,Tunnel,Part of MBone,R,R,R,H,R,H,R,R,H,27,Multicast OSPF (MOSPF),MOSPF,是,OSPF,的扩充，以便支持组播,路由器（,Multicast-capable routers,）广播链路状态广告（,link state routing advertisements,）,链路状态报文中包含有组播地址（,Link-state packets,include multicast group addresses,to which local members have joined,）,一个路由器接到组播数据包时就计算（应用,Dijkstra,算法）到目标节点的最短路径以便形成组播树,28,Source 1,Receiver 1,Receiver 2,MOSPF: Example,Z,W,Q,T,29,Source 1,Receiver 1,Receiver 2,Link Failure/Topology Change,Z,W,Q,T,X,30,Source 1,Receiver 1,Receiver 2,Group Membership Change,Z,W,Q,T,Receiver 3,31,共享树,vs.,源基树,源基树（,Source-based trees,）,为每一个源端建立一个组播树,每个路由器维护,(S,G),状态（,state,）的组播路由表,Eg: DVMRP, MOSPF, PIM-DM, PIM-SM,共享树（,Shared trees,）,所有成员共享一个组播树,无论哪个源端都是用同一个组播树传输数据,路由器只需要维护,(*,G),状态（,(*,G) state at intermediate routers,）,Eg: CBT, PIM-SM,32,源基树（,Source-based Trees,）,Router,Source,Receiver,S,R,R,R,R,R,S,S,33,共享树（,Shared Tree,）,RP,Router,Source,Receiver,S,S,S,R,R,R,R,R,34,Shared vs. Source-Based Trees,Source-based trees,Shortest path trees low delay, better load distribution,More state at routers (per-source state),在密集模式的网络环境下比较有效（,Efficient in,dense-area,multicast,）,Shared trees,Higher delay (bounded by factor of 2), traffic concentration,Choice of core affects efficiency,Per-group state at routers,比较适合稀疏模式的网络环境（,Efficient for,sparse-area,multicast,）,35,Core-based Routing Protocols,Specify “meeting place” aka “core” or “rendezvous point (RP)”,Sources send initial packets to core,Receivers join group at core,Requires mapping between multicast group address and “meeting place”,Examples:,CBT, PIM-SM,36,协议独立组播路由协议：,Protocol Independent Multicast (PIM),Support for both shared and per-source trees,Dense mode (per-source tree),Similar to DVMRP,Sparse mode (shared tree),Core = rendezvous point (RP),Independent of unicast routing protocol,Just uses unicast forwarding table,37,PIM Protocol Overview,Basic protocol steps,Routers with local members Join toward Rendezvous Point (RP) to join shared tree,Routers with local sources encapsulate data in Register messages to RP,Routers with local members may initiate data-driven switch to source-specific shortest path trees,PIM v.2 Specification (RFC2362),38,Source 1,Receiver 1,Receiver 2,PIM Example: Build Shared Tree,(*,G),Receiver 3,(*,G),(*,G),(*,G),(*,G),(*,G),Join messagetoward RP,Shared tree after,R1,R2 join,RP,39,Source 1,Receiver 1,Receiver 2,Data Encapsulated in Register,(*,G),Receiver 3,(*,G),(*,G),(*,G),(*,G),(*,G),Unicast encapsulated data packet to RP in,Register,RP,RP de-capsulates, forwards down shared tree,40,Source 1,Receiver 1,Receiver 2,RP Send Join to High Rate Source,Receiver 3,(S1,G),RP,Join messagetoward S1,Shared tree,41,Source 1,Receiver 1,Receiver 2,Build,Source-Specific,Distribution Tree,Receiver 3,Join messages,Shared Tree,RP,Build source-specific tree for high data rate source,(S1,G),(*,G),(S1, G),(S1,G),(*,G),(S1,G),(*,G),42,Thank you !,43,