S. Katti, H. Rahuk, W. Hu, D. Katabi, M. Medard, J. Crowcroft, "XORs in the Air: Practical Wireless Network Coding," ACM SIGCOMM Conference, (September 2006).
This paper talks about COPE, a new architecture for wireless mesh networks. The idea is to cleverly use network coding to mix packets from different sources in order to increase the information content in each transmission. COPEs design is based on 3 main principles:
- It effectively utilize the broadcast nature of wireless channel to exploit opportunistic learning. All the nodes are kept in promiscuous mode and passively listen for broadcast packets.
- Use of Network Coding in a practical scenario. The sender effectively XORs various packets together and depends on neighbour nodes to decode individual packets.
- Using reception reports by which adjacent nodes can find out about the packets that their neighboring node contains.
COPE definitely performed well with UDP, resulting in a throughput increase of upto 3-4 times, however, it didn't result in any significant improvement with TCP. This difference in the behavior of COPE based on protocols and topolology raises questions about the practicality of its actual deployment. However, due credit goes to the authors for cleverly using network coding for increasing throughput. It would be interested to know about some techniques that may have attempted to combine COPE with ExOR in a more conclusive manner.
I agree completely. A clever idea, but unclear how generally applicable it is. The lack of support for TCP is problematic, but perhaps alternative transports, like reliable UDP, are perhaps a better match. Managing the packet cache is also an issue.
ReplyDelete