Caching Techniques to Improve Latency in Serverless Architectures

TL;DR

This paper analyzes latency issues in serverless architectures, especially with AWS Lambda, and proposes caching strategies that significantly improve response times compared to traditional VM setups.

Contribution

It provides a comparative analysis of serverless versus VM architectures and introduces caching techniques to reduce latency in serverless environments.

Findings

Database access latency in serverless is 14 times higher than in VM-based setups.

Caching strategies can significantly reduce serverless response times.

Performance improvements depend on the caching method used.

Abstract

Serverless computing has gained a significant traction in recent times because of its simplicity of development, deployment and fine-grained billing. However, while implementing complex services comprising databases, file stores, or more than one serverless function, the performance in terms of latency of serving requests often degrades severely. In this work, we analyze different serverless architectures with AWS Lambda services and compare their performance in terms of latency with a traditional virtual machine (VM) based approach. We observe that database access latency in serverless architecture is almost 14 times than that in VM based setup. Further, we introduce some caching strategies which can improve the response time significantly, and compare their performance.