From: Michelle Liu (liujing@u.washington.edu)
Date: Sun Nov 14 2004 - 15:21:41 PST
Review of "Internet Indirection Infrastructure"
Jing Liu
This paper presents an overlay-based Internet indirection infrastructure (i3) that offers a rendezrous-based communication abstraction. i3 decouples the act of sending from the act of receiving so that it can support a variety of fundamental communication services, such as multicast, anycast, and host mobility.
The author compares i3 with IP multicast. IP multicast offers a receiver a binary decision of whether or not to receive packets sent to that group. It is up to the multicast infrastructure to build efficient delivery trees. While the i3 abstraction inserts a trigger, which is more flexible in that it allows receivers to control the routing of the packet. The advantages are: first, a variety of application services such as mobility and anycast can be built out of this basic service model. Second, the end hosts are responsible for efficient tree construction. This allows the infrastructure to remain simple, robust, and scalable.
The main design of i3 includes the following. Rendezvous-based communication and use of trigger enable efficiency of building delivery trees. In order to provide reliability of communication, Chord lookup protocol is used such that multiple servers are responsible for one identifier. Moreover, each receiver maintains a backup trigger or the overlay network itself replicate the triggers. With regard to scalability issue, hierarchy of triggers is constructed and maintained either cooperatively by the members of the multicast group, or by a third party provider. Stack of identifiers are proposed to implement the hierarchy architecture. As for routing efficiency, this hierarchy architecture is also helpful and moreover, the receivers will choose triggers that are located on nearby servers in order to avoid triangle routing.
One problem is that in order to provide robustness and reliability, triggers will be replicated and stored in the network. This will require additional storage overhead. Especially, how much the replica should be enough to keep reliability and will that affect scalability?
This is a very interesting paper considering today's application requirements. People are more and more concerned with mobility and multicast. If more and more such applications could be built upon the i3 infrastructure, the implementation of i3 is very meaningful.
This archive was generated by hypermail 2.1.6 : Sun Nov 14 2004 - 15:22:01 PST