On Composition and Implementation of Sequential Consistency (Extended Version)

TL;DR

This paper extends the understanding of sequential consistency in distributed systems, proposing an algorithm for crash-prone asynchronous environments and analyzing when waiting is necessary for consistency guarantees.

Contribution

It introduces a new distributed algorithm for sequentially consistent shared memory with snapshots in asynchronous crash-prone systems and analyzes its composability and waiting conditions.

Findings

Waiting is only necessary after a write on the same process before a read.

Sequential consistency can be composable in certain practical scenarios.

The proposed algorithm works with less than half of processes crashing.

Abstract

It has been proved that to implement a linearizable shared memory in synchronous message-passing systems it is necessary to wait for a time proportional to the uncertainty in the latency of the network for both read and write operations, while waiting during read or during write operations is sufficient for sequential consistency. This paper extends this result to crash-prone asynchronous systems. We propose a distributed algorithm that builds a sequentially consistent shared memory abstraction with snapshot on top of an asynchronous message-passing system where less than half of the processes may crash. We prove that it is only necessary to wait when a read/snapshot is immediately preceded by a write on the same process. We also show that sequential consistency is composable in some cases commonly encountered: 1) objects that would be linearizable if they were implemented on top of a linearizable memory become sequentially consistent when implemented on top of a sequential memory while remaining composable and 2) in round-based algorithms, where each object is only accessed within one round.