review

From: Jenny Liu (jen@cs.washington.edu)
Date: Wed Nov 10 2004 - 04:29:21 PST

  • Next message: T Scott Saponas: "Review of The PIM Architecture for Wide-Area Multicast Routing"

    "The PIM Architecture for Wide Area Multicast Routing" presents an
    approach and protocol for sending multicast data over a wide area,
    particularly when receivers are sparsely distributed. PIM combines
    ideas from the CBT protocol (namely shared trees) and the SPT (shortest
    path tree) from the traditional IP multicast model by allowing routers
    with receivers as local members along PIM's RP (shared) tree to choose
    between a shared tree and a source-specific tree.

    By giving the choice of multicast districution tree type to routers
    directly connected to receiving members, PIM provides for
    end-point-based choice in the type of tree used for multicasting. The
    choice of tree type can even be changed on the fly, allowing receivers
    (or their first hop routers) to choose what's best for them depending on
    the specific multicast application. However, the question of how first
    hop routers know whether they want RP or SPT paths within their networks
    still remains. It seems that in the end, a human will have to decide on
    (at the least) some parameters to govern the state of the tree choice
    within each network, and it's not clear how those parameters are to be
    chosen.

    The paper points out that multicasting on RP and SPT type trees behaves
    well under different circumstances. How often do those circumstances
    change for the same set of senders and receivers? It's not clear that
    the flexibility provided by being able to switch from one type of tree
    to the other is actually very useful.


  • Next message: T Scott Saponas: "Review of The PIM Architecture for Wide-Area Multicast Routing"

    This archive was generated by hypermail 2.1.6 : Wed Nov 10 2004 - 04:29:19 PST