Review - Chord: A Scalable P2P Lookup Protocol for Internet Applications

This paper introduces Chord, a decentralized and robust lookup system
for large P2P systems. For a Chord network of N nodes, lookups can be
resolved with certainty in O(log N) messages and each node only needs to
maintain O(log N) routing information to other nodes. Chord performs
lookups based on a key and it resolves the lookup by returning the IP
address of a node containing the key. Chord is simple and provably correct.

The strength of this paper lies in its Chord design of course, which is
very elegant and promising. Above all, I liked the paper's clarity which
makes reading and understanding easy. The use of examples in the paper
to explain the algorithm provides very good insight on Chord's
algorithm. Most assumptions, pitfalls and realistic scenarios are put
under scrutiny and are explained, which I believe is important.

The weakness of this paper, in my opinion, is that it assumed that all
values are kept on the network at all times; unless I got it wrong. It
is unclear to me how Chord would perform in an file-sharing P2P
application where values (files) are stored on the node and these
disappear when a node leaves the network. It is also unclear how they
will deal with nodes joining the network with lots of files (much more
than the other nodes) and load balance the Chord network subsequently.

I believe that Chord is very promising as per the potential application
of Chord outlined in the paper. For instance, there is always fear that
one could bring down the Internet by attacking the root DNS servers. I
believe Chord could be an ideal candidate to replace DNS and make it
more robust.
Received on Mon Nov 07 2005 - 10:26:40 EST