A Study of Synchronization Methods for Concurrent Size

TL;DR

This paper compares various synchronization techniques for maintaining collection sizes in concurrent data structures, showing that the optimal method depends on contention levels, with no single best approach for all scenarios.

Contribution

It provides a comprehensive evaluation of synchronization methods for concurrent size operations, highlighting their performance trade-offs under different contention conditions.

Findings

Optimistic and lock-based methods excel under low contention.

Handshake and wait-free approaches perform better under high contention.

No single synchronization method is optimal for all scenarios.

Abstract

The size of collections, maps, and data structures in general, constitutes a fundamental property. An implementation of the size method is required in most programming environments. Nevertheless, in a concurrent environment, integrating a linearizable concurrent size introduces a noticeable overhead on all operations of the data structure, even when the size method is not invoked during the execution. In this work we present a study of synchronization methods in an attempt to improve the performance of the data structure. In particular, we study a handshake technique that is commonly used with concurrent garbage collection, an optimistic technique, and a lock-based technique. Evaluation against the state-of-the-art size methodology demonstrates that the overhead can be significantly reduced by selecting the appropriate synchronization approach, but there is no one-size-fits-all method. Different scenarios call for different synchronization methods, as rigorously shown in this study. Nevertheless, our findings align with general trends in concurrent computing. In scenarios characterized by low contention, optimistic and lock-based approaches work best, whereas under high contention, the most effective solutions are the handshake approach and the wait-free approach.