Year: 2024
Type: article
Abstract: Serverless computing has emerged as a dominant cloud paradigm due to its automatic scaling, event-driven execution model, and pay-per-use cost structure; however, these advantages are often constrained by performance unpredictability, cold-start delays, and suboptimal resource allocation strategies [25],[24]. Recent studies show that as applications grow more latency-sensitive and distributed, traditional serverless platforms struggle to balance cost efficiency and high performance, especially under bursty or heterogeneous workloads [23],[26]. This paper investigates optimization approaches spanning cold-start reduction, adaptive memory tuning, workflow orchestration improvements, and predictive scaling techniques. Drawing on empirical findings that link runtime selection and concurrency management to significant latency reductions [4],[28], this study develops an integrated optimization framework designed to enhance both execution performance and cost efficiency in Function-as-a-Service (FaaS) environments. Experimental evaluation demonstrates that combining lightweight runtime containers, machine-learning-guided resource allocation, and improved state management can substantially reduce end-to-end latency while lowering operational cost. The results contribute to ongoing efforts to make serverless platforms more predictable, scalable, and suitable for enterprise-grade workloads.
Author: Sanketh Nelavelli
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W7139024909