Towards Serverless Optimization with In-place Scaling

TL;DR

This paper investigates how in-place scaling in Kubernetes v1.27 can reduce cold start latency in serverless computing, demonstrating significant improvements in request response times across different workloads.

Contribution

It provides an empirical analysis of in-place scaling's impact on serverless latency, highlighting its potential to mitigate cold start delays.

Findings

Latency reductions of 1.16 to 18.15 times across workloads

In-place scaling improves serverless responsiveness

Experimental results validate the effectiveness of in-place scaling

Abstract

Serverless computing has gained popularity due to its cost efficiency, ease of deployment, and enhanced scalability. However, in serverless environments, servers are initiated only after receiving a request, leading to increased response times. This delay is commonly known as the cold start problem. In this study, we explore the in-place scaling feature released in Kubernetes v1.27 and examine its impact on serverless computing. Our experimental results reveal improvements in request latency, with reductions ranging from 1.16 to 18.15 times across various workloads when compared to traditional cold policy.