Hybrid Stochastic-Dynamic Framework for Latency and Energy-Aware Task Offloading in Edge Computing

Abstract

A system that enables easy, network-on-demand access to a shared and customizable pool of computer resources is cloud computing. Resource provisioning, a key component of Infrastructure as a Service in the cloud, encounters issues with pricing ambiguity, heterogeneity, availability, and demand consideration. To solve the problems, approximation dynamic programming and stochastic programming (SP) are offered, which increase the cloud computing framework's four main uncertainties: price, heterogeneity, demand, and availability. A tree is built to reduce the space in SPDP, a multi-stage paradigm for dynamic programming that takes uncertainty into account in the underlying probability space. The reduction of space successfully promotes exploration, resource over- and under-provisioning, and significant profit-enhancement.The suggested method exemplifies responsive pricing research that is advantageous to both users and suppliers. As compared to other accessible algorithms, the SPDP provides promising results, according to the results analysis.