Fault Tolerant scheduling in Cloud Data Centers for improving Resource Utilization

Abstract

Most of the existing fault tolerance mechanism focuses on creating replicas of virtual newlinemachines to replace in case of virtual machine failure which results in tremendous wastage of newlineresources rather than focusing on earlier predicting the failure in advance. In case of VM newlinefailure, some of the existing algorithms focus on migration without considering the resource newlineconstraints or availability of the target server. Most of the algorithms are single objective newlinebase like fault tolerance or only predict the failure or the migrating with no resource newlineutilization. In this work we try to propose an optimal solution from the early prediction of newlinefailure till the optimal migration. newlineEarly Prediction of failure in virtual machines plays a vital role because of several issues like newlinewastage of resources, energy and cost. Reliability of VMs has always been a challenge in a newlinecloud environment. A fault tolerance (FT) framework that performs environmental newlinemonitoring, event logging, parallel job monitoring and resource monitoring to analyze the newlinevirtual machine reliability and to perform fault tolerance service are very much required to newlinehandle these challenges. As a part of fault tolerance mechanism there is a thorough necessity newlinefor providing preventive solutions to have continuity of services. Hence the proactive failure newlineprediction of Virtual Machine (VMs) needs to be focused and also to be improved. It is newlinemainly required to reduce the down time and cope up the scalability issues. The VM newlinemigration process can influence the execution of applications unless it is supported by better newlineoptimization techniques. A multi-objective optimization (MOO) model has been proposed in newlinethis work to address this issue. A migration Algorithm has been used to migrate the predicted newlineto be failed resources from one VM to another VM safely, and by using the compression newlinealgorithm, the down time and the compressed time have reduced and resource utilization has newlineincreased. newlineIn reality and supported by literature review fault tolerance is a key challenge in cloud data newlinecenter resources maintenance and moreover maintaining multiple servers leads to server newlinesprawling. At present, to provide uninterrupted services, multiple replicas of VMs are created newlineto face this challenge. However, this will result in serious problems like wastage of resources, newlineconsumption of energy, and rising costs. Based on the famous saying Prevention is better newlinethan cure this work has been undertaken to predict the failure in advance and also optimize newlinethe various key resources. newlineThis framework is to reduce the wastage of resources and predict failure earlier, thereby newlineminimizing costs, saving time and without violating the SLAs. newline