Energy and cost-aware workload scheduler for heterogeneous cloud platform

Abstract

Parallel scientific workloads, often represented as directed acyclic graphs (DAGs), consist of interdependent tasks that require significant data exchange and are executed on distributed clusters. The communication overhead between tasks running on different nodes can lead to substantial increases in makespan, energy usage, and monetary costs. Therefore, there is potential to balance communication and computation to reduce these costs. In this paper, we introduce an energy and cost-aware workload scheduler (ECAWS) tailored for executing parallel scientific workloads, generated by the internet of things (IoT), in a heterogeneous cloud environment. The performance of the proposed ECAWS model is evaluated against existing models using the Inspiral scientific workload. Results indicate that ECAWS outperforms other models in reducing makespan, costs, and energy consumption.