Serverless Cloud Computing: A Comparative Analysis of Performance, Cost, and Developer Experiences in Container-Level Services

Abstract

Serverless cloud computing is a subset of cloud computing considerably adopted to build modern web applications, while the underlying server and infrastructure management duties are abstracted from customers to the cloud vendors. In serverless computing, customers must pay for the runtime consumed by their services, but they are exempt from paying for the idle time. Prior to serverless containers, customers needed to provision, scale, and manage servers, which was a bottleneck for rapidly growing customer-facing applications where latency and scaling were a concern.

The viability of adopting a serverless platform for a web application regarding performance, cost, and developer experiences is studied in this thesis. Three serverless container-level services are employed in this study from AWS and GCP. The services include GCP Cloud Run, GKE AutoPilot, and AWS EKS with AWS Fargate. Platform as a Service (PaaS) underpins the former, and Container as a Service (CaaS) the remainder. A single-page web application was created to perform incremental and spike load tests on those services to assess the performance differences. Furthermore, the cost differences are compared and analyzed. Lastly, the final element considered while evaluating the developer experiences is the complexity of using the services during the project implementation.

Based on the results of this research, it was determined that PaaS-based solutions are a high-performing, affordable alternative for CaaS-based solutions in circumstances where high levels of traffic are periodically anticipated, but sporadic latency is never a concern. Given that this study has limitations, the author recommends additional research to strengthen it.