3.1 Dense Protocols

3.1 Dense Protocols Dense protocols assume a dense distribution of receivers exists throughout the domain, which means each subnet likely has at least one interested receiver for every active group. This assumption may be valid on enterprise networks where only a few groups are active and most of the subnets contain interested listeners. On networks where few or no prunes occur, dense protocols are actually more efficient than sparse protocols. However, on the Internet, where prunes are more prevalent, dense protocols are not well suited to interdomain deployment. Dense protocols follow a flood-and-prune model. To inform the routers of multicast sources, this traffic is initially broadcast throughout the domain. Upon first receiving traffic to a dense group on its interface closest to the source, a router forwards this traffic out all of its interfaces except the interface on which it received the data. Thus the IIF initially is the RPF interface toward the source, and the OIL contains all other interfaces. If traffic is received on the interface that is not the RPF interface toward the source, the traffic is discarded, and a Prune message is sent upstream. If a router has no interested receivers for the data (that is, its OIL becomes empty), it sends a Prune upstream. Periodic reflooding is used to refresh state. The primary benefit of dense protocols is simplicity. The flood-and-prune mechanism enables these protocols to easily build a multicast distribution tree rooted at the source. A source-based tree guarantees the shortest and most efficient path from source to receiver. The obvious limitation is scalability; any mechanism that relies on flooding across the entire network does not scale particularly well on the Internet. 3.1.1 DVMRP DVMRP is the multicast routing protocol first used to support the MBone. It performs the standard flood-and-prune behavior common to dense protocols. It also implements a separate routing protocol used to build the routing tables on which RPF checks are performed. As its name suggests, DVMRP has a distance-vector routing protocol very similar to RIP. It has the same limitations found in other distance-vector protocols, which include slow convergence and limited metric (that is, hop count). Although most DVMRP deployments have been replaced by PIM-SM, DVMRP can still be found on networks with legacy equipment, such as dialup RASs. Most RASs made today support either PIM-SM or IGMP proxying. 3.1.2 PIM-DM PIM-DM implements the same flood-and-prune mechanism mentioned previously and is quite similar to DVMRP. The primary difference between DVMRP and PIM-DM is that the latter aptly named protocol introduces the concept of protocol independence. PIM, in both dense and sparse modes of operation, can use the routing table populated by any underlying unicast routing protocol to perform RPF checks. This ability to use any underlying unicast protocol was seen by ISPs to be a significant enhancement because they did not want to manage a separate routing protocol just for RPF (ironically, MBGP and M-ISIS were later used to do just that). PIM-DM can consult the unicast routing table, populated by OSPF, IS-IS, BGP, and so on, or it can be configured to use a multicast RPF table populated by MBGP or M-ISIS when performing RPF checks.