Previous | Table of Contents | Next |
When a host receives a query, it can start a report delay timer rather than sending reports immediately for each of its group members on the network interface of the query. Timers are set to different, randomly chosen values between zero and D seconds. A report is generated for the corresponding host group when a timer expires. Thus, reports occur over a D-second interval rather than concurrently.
To ensure that other members of the same group overhear the report, it is sent with an IP destination address equal to the host group address being reported and with an IP time-to-live of 1. If a host hears a report for its group on the network, the host stops its own timer and does not generate a report for that group. Thus, usually only one report is generated for each group on the network, by the member host whose delay timer expires first. The multicast routers receive all IP multicast datagram, therefore they need not be addressed explicitly. The routers need not know which hosts belong to a group, only that at least one host belongs to a group on a particular network.
Multicast routers send queries periodically to refresh their knowledge about memberships on a particular network. Queries are usually sent at intervals of approximately one minute to minimize overhead traffic on the network. However, when a multicast router starts up, it may often issue several closely spaced queries to obtain knowledge of local group membership quickly.
Similarly, when a host joins a new group, it issues a report immediately without waiting for a query in case it is the first member of the group on the network.
New members notify the nearest router of their presence in a group almost instantly, whereas notification of leaving a group depends on query timeout, which can be a minute or more. A new version of IGMP is being proposed that provides for explicit host leave group reports, to speed up the notification to routers of hosts leaving a group.
The Mbone of the Internet is based on multicast IP and covers 3,000 subnetworks. It is predicted that the Internet will be fully multicast IP enabled soon.
The Mbone uses computers to provide the multicast routing capability, and multicast packets are tunneled through unicast links to tie together the islands of multicast networks. It is used mostly by Internet researchers for videoconferencing and dataconferencing.
Two applications used on the Mbone are based on the ALF lightweight session model originated by Internet researchers. ALF dictates that the best way to meet diverse multicast application requirements is to leave as much flexibility as possible to the application in the application layer. This means that connectivity is viewed as unreliable, using the UDP connectionless transport layer.
Two prominent ALF model applications come from Lawrence B. Livermore Laboratory, which provides the Visual/Audio Tool in multimedia videoconferencing. In addition, the Mbone enables a whiteboard application that lets users of a multicast group electronically write on a whiteboard. All members of the group can see what each member has written. Whiteboarding is essentially an electronic dataconference that allows group members to brainstorm visually.
Other videoconferencing tools that provide use over the Mbone are being used in some universities; however, they are not yet commercially available. The Mbone is still in the research stage.4
Multicast routing protocols are required to perform optimal routing through router networks, just as unicast routing protocols such as IGMP, OSPF, and RIP are needed to perform optimal unicast routing in router networks. Multicast routing protocols should efficiently minimize the necessary traffic for routing multicast data in the network or internetwork.
There are three predominant multicast routing protocols: DVMRP, Multicast OSPF, and PIM.
Exhibit 6. DVMRP Flooding.
Exhibit 7. Sparse Mode Topology Showing Rendezvous Points.
Previous | Table of Contents | Next |