An Efficient Renewable Energy-Based Scheduling Algorithm for Cloud Computing

Abstract

The global growth of cloud computing services is witnessing a continuous surge, starting from storing data to sharing information with others. It makes cloud service providers (CSPs) efficiently utilize the existing resources of datacenters to increase adaptability and minimize the unexpected expansion of datacenters. These datacenters consume enormous amounts of energy generated using fossil fuels (i.e., non-renewable energy (NRE) sources), and emit a substantial amount of carbon footprint and heat. It drastically impacts the environment. As a result, CSPs are pledged to decarbonize the datacenters by adopting renewable energy (RE) sources, such as solar, wind, hydro and biomass. However, these CSPs have not completely ditched fossil fuels as RE sources are subjected to inconsistent atmospheric conditions. Recent studies have suggested using both NRE and RE sources by the CSPs to meet user requirements. However, these studies have not considered flexible duration, nodes and utilization of the user requests (URs) with respect to datacenters. Therefore, we consider these URs' properties and propose a RE-based scheduling algorithm (RESA) to efficiently assign the URs to the datacenters. The proposed algorithm determines both the earliest completion time and energy cost, and takes their linear combination to decide a suitable datacenter for the URs. We conduct extensive simulations by taking 1000 to 16000 URs and 20 to 60 datacenters. Our simulation results are compared with other algorithms, namely roundrobin (RR) and random, which show that RESA is able to reduce the overall completion time (i.e., makespan (M )), energy consumption (EC), overall cost (OC) and the number of used RE (|URE|) resources.

Analyzing Power Decisions in Data Center Powered by Renewable Sources

“…Several works pursue ways to deal with RES uncertainties. However, they introduce, in many cases, grid (brown) connections or just one level of management (online or offline) [10], [11], [12], [13], [14], [15]. In work [14], the authors created an offline optimization framework using a model to capture the randomness of the RES.…”

“…SMM is similar to GA but uses frames to compare and improve the best solution. In [11], the authors create an online heuristic to assign the user requests to the data centers. Their objective is to reduce the makespan, energy consumption, and overall cost.…”

Evaluation of Heuristics to Manage a Data Center Under Power Constraints

“…The main objective is to reduce energy costs. In [7], the authors create an algorithm to efficiently assign the user requests to the data centers, reducing the makespan, energy consumption, and overall cost. [8] designed an algorithm to group similar service-level-objective (SLO) jobs in the same physical machine group powered by renewable sources.…”

Energy‐aware autoscaling for scientific workflow in cloud environment

“…Botero et al 30 described VN embedded energy-efficient situation and developed a MIP model to accomplish energy-aware VN embedding for this aspect. Nayak et al 9,31,32 focused on renewable energy based job scheduling technique. In Reference 9 authors proposed scheduling algorithm for the determination of suitable datacenter to the user requests based on the linear combination of calculated earliest completion time and energy cost.…”

“…Recently cloud infrastructures are also being evaluated as an execution platform for workflows. 8,9 Cloud computing represents a new way of thinking about how to deploy and execute scientific workflows.…”