Improved Environmental Adaptation Method for Scheduling Workflows in Cloud

Abstract

Cloud users are expanding at rapid rate which forces the cloud data centres execute billions of commands each second. A random user request must be planned and processed on the workflow without knowing the sequence of future requests. This makes workflow scheduling on distributed environment as NP-hard problem. In this work we present an optimization-based scheduling approach that responds to cloud’s dynamic nature. The suggested technique derives from the Environmental Adaptation Method (EAM), an evolutionary algorithm established to handle optimization problems. After EAM’s original proposal, multiple better versions were made to fix inherent issues. Most of the revised algorithms performed well in lower dimensions but degraded performance is seen in higher ones. Most of these methods were binary encoded, which poses issues for real-valued parameters owing to conversion cost. Improved Environmental Adaptation Method with Real parameters (IEAM-R) was presented to deal with real valued problems to increase IEAM’s convergence rate. IEAM-R performs effectively on lower-dimensional benchmark functions, but not on larger dimensions. We changed IEAM-R and created a new algorithm to increase the diversity of solutions in higher dimensions. Exploration and exploitation must be redesigned to im-prove convergence rate. On all 24 benchmark functions, the proposed modified optimization algorithm with fine-tuned operators, outperforms its predecessors and other state-of-the-art algorithms. The technique is then used to the workflow scheduling issue in cloud computing, where it reduces the overall cost of cloud operation as compared to other heuristic and metaheuristic approaches.