Thursday, October 30, 2008

reading on "Chord"

Summary: This paper presents "Chord" using consistent hashing which is designed to let nodes enter and leave the network with minimal disruption. The chord protocol is described clearly in the paper in which some important features like simplicity, provable correctness and provable performance are also discussed. The authors believe that Chord will become a valuable component for peer-to-peer systems.
Background: consistent hashing
Discussion:
1, i feel a little bit lost about Chord's Scalable Key Location. How Chord makes it?
2, in another paper the authors talked about DHT, but in Chord it doesn't talk anything about DHT. Cause i am not familiar with DHT, my question is: Can we say DHT is a consistent hashing technique?
3, is Chord deployed somewhere in current peer-to-peer systems? Based on this paper, Chord should attract lots of attention from the peer-to-peer systems' designers.

Monday, October 27, 2008

Reading On "RON"

Summary: This paper presents RON- resilient overlay network which tights the applications and routing to let end-to-end communications be unaffected in face of path outlets. RON is a novel way to deal with the underlying network protocol's problems, especially BGP's long convergence time caused by link failures. 
Background: BGP
Discussion:
This is a very good paper which has been cited by 1185 times. I think the following reasons which let this paper be so popular:
1, No matter RON is practical or not,  the idea in it is quite new that tries to solve underlying network problems in the end user perspective. (a new story of End to End argument?)
2, Real data from real world. The author deployed 16 RON nodes in the Internet and got convincing results to show the RON's effectiveness.
3, Authors themselves has realized the RON's problems and explained them clearly in paper.

Thursday, October 9, 2008

reading on "ExOR: Opportunistic Multi-Hop Routing for Wireless Networks"

Summary: this paper presents ExOR which basically lets one node to send packets to several nodes and then chooses one of these nodes continues to send to several nodes till the packets reach the destination.  There are several challenges in ExOR. Authors built a packet format to solve these challenges. Through experiments on Roofnet, ExOR makes a big improvement on throughput than traditional routing scheme.
Background: ETX
Discussion:
this is a good paper and really smart idea. 
1, is ExOR really practical? 
2, i do think there are too much stuff in the ExOR packet header. Is it a good solution to augment the packet header?

Tuesday, October 7, 2008

reading on "A high-throughput path metric for multi-hop wireless routing"

Summary: this paper presents a new path metric called Estimated Transmission Count(ETX) which can be used to find higher throughput paths than using minimum-hop-count. The author briefly described DSDV and DSR and made them be capable to incorporate ETX. 
Background:
Discussion:
1, what is motivation for finding the higher throughput paths?

Wednesday, October 1, 2008

reading on "Modeling Wireless Links for Transport Protocols"

summary:this paper argues that because of the inaccurate modeling for the link layer, the studies on the transport layer have more or less problems. The authors briefly reviewed the different link characteristics, topologies , traffic pattern and performance metrics of cellular, WLAN and satellite networks. The paper lists all the aspect which should be taken consideration to model link layer and gives the methods to model them. At the end of the paper, the authors talked about some future work.
background: AQM
Discussion:
1, what is AQM? it looks like AQM is an very important concept. Did we talked about it before?
2, it is true that wireless links become one important part of the current Internet. Should we really make TCP work under such a mixed circumstance. I do think connection-split is a good idea although proved to be useless in Hari's paper.
3, is this paper the first one to talk about the cross-layer communication? What is the biggest problem to introduce the cross-layer communication mechanism to current network protocol stack?