Using Weaker Consistency Models with Monitoring and Recovery for Improving Performance of Key-Value Stores

TL;DR

This paper presents a detect-rollback approach that allows existing algorithms designed for sequential consistency to operate efficiently under weaker consistency models like eventual consistency, by monitoring correctness predicates and rolling back when violations occur.

Contribution

The paper introduces a novel detect-rollback method enabling correct execution of algorithms under weaker consistency models by monitoring and restoring states, improving performance of key-value stores.

Findings

Eventual consistency with monitoring yields 50-80% higher throughput than sequential consistency.

Monitoring overhead is typically less than 4%.

Most violations are detected within 50 milliseconds in regional networks.

Abstract

Consistency properties provided by most key-value stores can be classified into sequential consistency and eventual consistency. The former is easier to program with but suffers from lower performance whereas the latter suffers from potential anomalies while providing higher performance. We focus on the problem of what a designer should do if he/she has an algorithm that works correctly with sequential consistency but is faced with an underlying key-value store that provides a weaker consistency. We propose a detect-rollback based approach: The designer identifies a correctness predicate, say $P$, and continues to run the protocol, as our system monitors $P$. If $P$ is violated (because of weaker consistency), the system rolls back and resumes the computation at a state where $P$ holds. We evaluate this approach with graph-based applications running on the Voldemort key-value store. Our experiments with deployment on Amazon AWS EC2 instances shows that using eventual consistency with monitoring can provide a $50\%$ -- $80\%$ increase in throughput when compared with sequential consistency. We also observe that the overhead of the monitoring itself was low (typically less than $4\%$) and the latency of detecting violations was small. In particular, in a scenario designed to intentionally cause a large number of violations, more than $99.9\%$ of violations were detected in less than 50 milliseconds in regional networks, and in less than 3 seconds in global networks. We find that for some applications, frequent rollback can cause the program using eventual consistency to effectively \textit{stall}. We propose alternate mechanisms for dealing with re-occurring rollbacks. Overall, for applications considered in this paper, we find that even with rollback, eventual consistency provides better performance than using sequential consistency.