Parsley's Group Size Study

TL;DR

This paper presents Parsley's resilient group-based DHT with proactive size management, analyzing how its parameters influence stability, performance, and scalability under churn through systematic overlay characterization.

Contribution

It introduces a novel group size management approach with soft and hard limits, and provides a systematic analysis of parameter effects on system stability and scalability.

Findings

Soft limits improve stability under churn.

Parameter choices significantly affect performance.

Systematic analysis guides optimal configuration.

Abstract

Parsley is a resilient group-based Distributed Hash Table that incorporates a preemptive peer relocation technique and a dynamic data sharding mechanism to enhance robustness and balance. In addition to the hard limits on group size, defined by minimum and maximum thresholds, Parsley introduces two soft limits that define a target interval for maintaining stable group sizes. These soft boundaries allow the overlay to take proactive measures to prevent violations of the hard limits, improving system stability under churn. This work provides an in-depth analysis of the rationale behind the parameter values adopted for Parsley's evaluation. Unlike related systems, which specify group size limits without justification, we conduct a systematic overlay characterization study to understand the effects of these parameters on performance and scalability. The study examines topology operations, the behavior of large groups, and the overall trade-offs observed, offering a grounded explanation for the chosen configuration values.