A Fault Tolerant Elastic Resource Management Framework Toward High Availability of Cloud Services

Cloud computing has become inevitable for every digital service which has\nexponentially increased its usage. However, a tremendous surge in cloud\nresource demand stave off service availability resulting into outages,\nperformance degradation, load imbalance, and excessive power-consumption. The\nexisting approaches mainly attempt to address the problem by using multi-cloud\nand running multiple replicas of a virtual machine (VM) which accounts for high\noperational-cost. This paper proposes a Fault Tolerant Elastic Resource\nManagement (FT-ERM) framework that addresses aforementioned problem from a\ndifferent perspective by inducing high-availability in servers and VMs.\nSpecifically, (1) an online failure predictor is developed to anticipate\nfailure-prone VMs based on predicted resource contention; (2) the operational\nstatus of server is monitored with the help of power analyser, resource\nestimator and thermal analyser to identify any failure due to overloading and\noverheating of servers proactively; and (3) failure-prone VMs are assigned to\nproposed fault-tolerance unit composed of decision matrix and safe box to\ntrigger VM migration and handle any outage beforehand while maintaining desired\nlevel of availability for cloud users. The proposed framework is evaluated and\ncompared against state-of-the-arts by executing experiments using two\nreal-world datasets. FT-ERM improved the availability of the services up to\n34.47% and scales down VM-migration and power-consumption up to 88.6% and\n62.4%, respectively over without FT-ERM approach.\n