In P2P Knowledge Management Workshop at MobiQuitous, July 2005.
The scalability of publish/subscribe (pub/sub) systems and distributed hash tables (DHTs) have been extensively studied in the literature. However, less well-known are properties of the pub/sub model and DHTs that make them suitable for small networks. This paper articulates these benefits, and evaluates the performance of a DHT-based pub/sub implementation in small-scale networks. We find that a fundamental assumption of DHT-based data management applications is violated in small networks, and this makes the pub/sub implementation exhibit poor load balance under certain workloads. This work illustrates that data management applications that scale in large networks may not scale in small networks.
Readers who enjoyed the above work, may also like the following: