(no subject)

This paper has described in detail SRM (Scalable Reliable Multicast),
a reliable multicast framework for light-weight sessions and
application level framing. The SRM framework meets a minimal
reliability definition of delivering all data to all group members,
deferring more advanced functionality, when needed, to individual
applications. The algorithms of this framework are efficient, robust,
and scale well to both very large networks and very large sessions.
Multicast, by its very nature, is not a connection-oriented
mechanism, so protocols such as TCP, which allows for retransmission
of missing packets, are not appropriate. For applications such as
streaming audio and video, the occasional dropped packet is not a
problem. But for distribution of critical data, a mechanism is
required for requesting retransmission.
This paper mainly focuses on the reliable multicast framework. There
are already some proposed research work. This paper evolves the
principles of ALF and LWS to add a framework for SRM.
SRM is designed to meet only the minimal definition of reliable
multicast, and it's also heavily based on the group delivery model
that is the centerpiece of the IP multicast protocol. SRM further
enhances the multicast group concept by maximizing information and
data sharing among all the members, and strengthens the individuality
of membership by making each member responsible for its own correct
reception of all the data. Finally, SRM attempts to follow the core
design principles of TCP/IP.
This paper also focuses on the loss recovery, congestion control,
Request/repair algorithms for simple simple topologies. In SRM,
members who detect a loss wait a random time and then multicast their
repair request, to suppress requests form other members sharing that
loss. The interval over which the request timer is set is a function
of the member's estimated distance to the source of the packet. For a
chain topology, the essential feature of a loss recovery algorithm is
that the timer value is a function of distance. For a star topology,
the essential feature of the loss recovery algorithm is the
randomization used to reduce implosion. Request/repair algorithms in
a tree combine both the randomization and the setting of the timer as
a function of distance. The simple congestion control mechanism for
SRM would be for all members of the multicast group to assume a fixed
bandwidth constraint over the aggregate session.

This paper is expressed clearly and in detail. It is very solid with
analysis and simulation. The main contribution is that authors
describe the SRM framework in detail and discuss the wb instantiation
of this framework. It's research work with originality. However,
authors leave a lot of future work. Authors also investigate a
hierarchical approach for scalable session messages, where members in
a local area dynamically select one of the local members to be the
representative. The congestion control framework raises several
somewhat issues, such as how members determine this available
bandwidth; how to detect congestion or avoid potential congestion;
and given available bandwidth, which pice of data a member should
send first.
Received on Sat Oct 28 2006 - 16:48:56 EDT