Enhancing the efficiency and cost-reduction in serverless computing models on cloud platforms

Serverless computing has emerged as a transformative paradigm in cloud computing, enabling developers to focus on application development without the overhead of infrastructure management.This paper provides a comprehensive analysis of serverless computing, addressing its capabilities, challenges, and potential future directions.By synthesizing findings from 275 research articles, this study highlights the key benefits of serverless models, including cost reduction, enhanced scalability, and reduced latency.The research introduces innovative predictive models using multi-output regression, windowing techniques, and dimensionality reduction via Principal Component Analysis (PCA) to optimize function invocation predictability.Evaluations conducted on real-world datasets (e.g., Azure Functions) illustrate significant improvements in operational efficiency and temporal stability.Additionally, a taxonomy for assessing cost dynamics between serverless and traditional cloud approaches is proposed, supported by qualitative insights from industry experts.Furthermore, this study examines energy efficiency challenges in serverless data centers, proposing function-level power management systems and core-level scheduling policies that reduce energy consumption without compromising quality of service.For domain-specific applications, such as machine learning, the paper introduces MLLess, a Function-as-a-Service-based ML training prototype, achieving up to 15x speed improvements with cost efficiencies.These findings advance understanding and practical adoption of serverless computing, paving the way for future innovations in the field.