Generalize Synchronization Mechanism: Specification, Properties, Limits

TL;DR

This paper introduces a formal framework to analyze and compare synchronization mechanisms, highlighting their inherent trade-offs and limitations in achieving properties like consistency and availability.

Contribution

It proposes a generalized formal model for synchronization mechanisms, enabling better reasoning about their properties and trade-offs, extending concepts like CAP and ROLL theorems.

Findings

Identifies inherent trade-offs in synchronization mechanisms.

Provides a formal framework for analyzing synchronization properties.

Explores limitations preventing mechanisms from achieving all desired properties.

Abstract

Shared resources synchronization is a well studied problem, in both shared memory environment or distributed memory environment. Many synchronization mechanisms are proposed, with their own way to reach certain consistency level. This thesis further found that there is no perfect synchronization mechanism. Each of them has its properties at different level. For example, to enforce strong consistency, writers may loose writing freedom or it would take more time to coordinate. This thesis proposes a framework to generalize all synchronization mechanism in a formal way for better reasoning on properties, from the perspective of multi-writer to single-writer convergence. Therefore, limitations prevent a synchronization mechanism from achieving every property at its optimal level. CAP and ROLL were proposed in previous works to explain such. CAP theorem states that it can only achieve two of Consistency, Availability and Partition tolerance properties. ROLL Theorem uses a framework to model leaderless SMR protocol and states quorum size and fault tolerance are trading off. The thesis covers five properties in a more understandable way to analyze trade-offs and explore new mechanisms.