An analytical study on optimizing the lookup performance of distributed hash table systems under churn

Abstract

SUMMARYThe phenomenon of system churn degrades the lookup performance of distributed hash table (DHT) systems greatly. To handle the churn, a number of approaches have been proposed to date. However, there is a lack of theoretical analysis to direct how to make design choices under different churn rates and how to configure their parameters optimally. In this paper, we analytically study three important aspects on optimizing DHT lookup performance under churn, i.e. lookup strategy, lookup parallelism and lookup key replication. Our objective is to build a theoretical basis for designers to make better design choices in the future. We first compare the performance of two representative lookup strategies-recursive routing and iterative routing-and explore the existence of better alternatives. Then we study the effectiveness of lookup parallelism in systems with different churn rates and show how to select the optimal degree of parallelism. Owing to the importance of key replication on lookup performance, we also analyze the reliability of the replicated key under two different replication policies, and show how to perform proper configuration. Besides the analytical study, our results are also validated by simulation, and Kad is taken as a case to show the meaningfulness of our analysis.

A state management and persistency architecture for peer-to-peer massively multi-user virtual environments

“…From the figure it can also be seen that initial network size has a more pronounced effect on the expected object lifetime, when no repair is used than when repair is used. Wu, Tian and Ng that assumes object repair [123]. For this figure, a repair rate of µ = 1/180 and a required number of replicas of R = 10 were used.…”

“…This chapter mainly extends and improves upon the work by Wu, Tian and Ng [123]. The related work models three characteristics of distributed hash table (DHT) lookups under churn: expected lookup latency of various routing schemes, expected lookup overhead of various routing schemes and expected object lifetime.…”

“…In this chapter, we improve on the work by Wu, Tian and Ng [123] in two ways: firstly, the model is extended to take into account a finite network size, since there might not be sufficient nodes to replicate the data on when objects are stored or when the repair mechanism activates. Secondly, our model unifies the cases "with repair" and "without repair", to produce a single model where the effects of both might be evaluated.…”

“…In the work by Wu, Tian and Ng [123], an approximation is made for Pareto failure rate, given by θ = (α − 2)/β, where α and β are two parameters of the Pareto distribution. 1 During the calculation of expected object lifetimes, the expected lifetimes were found to be sensitive to small changes in the value of θ, which renders the approximation unusable for practical predictions using Pareto lifetimes.…”

An Improved Lookup Algorithm for Chord

“…Therefore, once the originator gets the expected result from one process, it can terminate another lookup process and avoid the remaining unnecessary costs for the overlay. Furthermore, iterative routing is robust to the high churn circumstance [7]. If there is no response after a timeout no less than the average round-trip time (RTT), the originator knows which hop failed to respond and will initiate a retry immediately.…”