Serverless Computing in Cloud-Based Application Development: Performance and Cost Analysis

Abstract

Serverless computing—primarily Function-as-a-Service (FaaS)—enables developers to deploy code without managing underlying infrastructure, charging only for actual execution time. This work evaluates performance and cost in serverless architectures relative to traditional and microservices-based deployments. We conducted latency, throughput, and scalability experiments across AWS Lambda, Azure Functions, and GCF, focusing on both batch workloads and intermittent, event-driven tasks. Cold-start delays were characterized, and mitigation techniques like pre-warming and function fusion were tested Cost modeling addressed pay-per-execution, memory allocation, and provider-specific pricing granularity . Our results show serverless offers significant cost advantages (30–60% lower TCO) for unpredictable or low-volume usage, but becomes less economical for continuous high-load applications due to execution overhead and resource limits . Performance analysis reveals cold starts add 100–500 ms latencies, impacting quality of service; edge-cloud hybrids and pre-warming strategies reduced this by ~50% Overall, while serverless offers scalability and reduced operational effort, challenges like latency spikes, debugging complexity, and vendor lock-in persist. We propose guidelines and a decision framework to assist architects in balancing performance, cost, and developer agility in choosing serverless versus container-based microservices.