Computational Intelligence Approach in Optimizing the Resource Allocation for Workflow Execution in the Cloud

Abstract

Abstract Virtual Machine (VM) assignment is an important phase during workflow execution in the cloud. Identifying a suitable type of VM for executing the workflow is a difficult problem. Cloud resource providers offer diverse categories of VMs to cater the needs of the users and encourage the users to select an appropriate type of VM to reduce the workflow execution time and cost. A suitable VM type can be selected only if the resource requirements of the workflow is known, which a difficult task for cloud users till now. Hence many users depend on general purpose VMs, rather than using a suitable type of VM for their application. This work proposes an intelligent recommendation system that helps the cloud users to select suitable type of VM for their application. The main objective of this work is to identify the type of the workflow using computational intelligence and to recommend a sutitable type of VM for execution. The system uses three supervised learning algorithms such as Probabilistic Neural network (PNN) , Deep Forward Neural Network (DFNN) and Naïve Bayes classier for workflow classification. The Bayes classifier is found be very accurate for workflow classification, when compared to the other algorithms. Also the relative optimality of the classifier is tested using popular workflow scheduling algorithms like ICPCP (IAAS Cloud Partial Critical Path algorithm), MER (Maximum Effective Reduction) and LBS (Level Based Scheduling). The results confirm that the workflow execution time and cost are reduced to a great extent, when executing a workflow using a suitable type of VM than with the general purpose VM.