Paper #: Week 4 Paper 1
Title: Chord: A scalable peer to peer lookup service for internet applications

(1) Problems
This paper tries to address the problem of storage and efficient lookup of a data item in a distributed network. The Chord protocol uses a consistent hashing system to store values. This is done in a decentralized manner. The paper presents an in depth explanation with supportive experiments and results of the Chord system and protocol.


(2) New Idea and Strengths
i) The most important strength of the protocol is the efficiency. A lookup is done in O(log N) [where N is the size of the network] messages to other nodes. This is a major improvement from the Gnutella algorithm which uses the flooding algorithm.
ii) Also, the distributed fingers hold information about only O(log N) neighbours.
iii) The cost of a new node joining or a node leaving the system is also not too expensive - 0(log2 N) messages.
iv) The above three properties are well thought and researched. They scale well with a large N which is very important in a peer to peer network which may have thousands of nodes.
v) The Chord protocol is an overlay system and is not an application on its own. A peer to peer application can be built using this. Thus, if the underlying system scales well, provides a best effort guarantee and is efficient amongst other properties, then the overlying application has inherent advantages.
vi) The paper is well written with suitable diagrams to explain concepts. Also, claims are proved via experiments and their results.


(3) Weaknesses and Extensions
i) Though the Chord protocol is efficient and scalable, it does not inherently provide anonymity, proper authentication or stronger consistency. These can be extensions to the system.
ii) One more weakness is that during a join operation the bootstrapping of a new node is handled offline.
In general I think the paper is well written. Its objective is clearly defined and it covers all aspects pretty solidly.


-- END --