A protocol for a distributed hash table (DHT) incurs communication costs to keep up with churn---changes in membership---in order to maintain its ability to route lookups efficiently. This paper formulates a unified framework for evaluating cost and performance. Communication costs are combined into a single cost measure (bytes), and performance benefits are reduced to a single latency measure. This approach correctly accounts for background maintenance traffic and timeouts during lookup due to stale routing data, and also correctly leaves open the possibility of different preferences in the tradeoff of lookup time versus communication cost. Using the unified framework, this paper analyzes the effects of DHT parameters on the performance of four protocols under churn.

Although structured peer-to-peer overlays are an increasingly popular area of research, ranges of performance both within a single overlay and between multiple overlays have yet to be fully examined. In particular, the effects of adverse conditions found in real-world networks on overlay performance have not been adequately quantified. In this paper, we present and analyze data extracted from the PlanetLab network, focusing mainly on the presence of churn and non-transitivity, and use this data to drive simulations of several structured peer-to-peer overlays. The protocols we use in our simulations are Chord, Tapestry, Kademlia, and Kelips. For each overlay, we simulate a range of different parameter settings, attempt to illustrate the tradeoffs inherent in parameter choices, and draw comparisons between the different protocols. We explore the performance of these overlays on a simulated PlanetLab topology, with and without the pathological conditions mentioned above. Our results indicate that non-transitivity can have a large effect on the performance of some overlays, while the amount of churn seen on PlanetLab tends to have a less significant impact.