HM-VM-AM – A HEURISTIC VS META-HEURISTIC APPROACHES FOR VMS ALLOCATION AND MIGRATION IN CLOUD COMPUTING

Abstract

Cloud computing survives on the basement of virtualization technology.  Implementation of efficient load balancing algorithms leads to proper distribution of dynamic workload among multiple nodes in cloud Data Centre (DC).  So that proper utilization of resources in DC is assured. VM load balancing is a NP-hard problem.  Live migration of VMs, become one of the necessary activities in DC for maintaining Service Level Agreements (SLA) and minimizing the number of servers in running state. Researchers are trying to use hybrid model of algorithms accompanying meta-heuristic vs. heuristic algorithms for VM placement and VM migration respectively to reduce the solution space and time cost.  The echolocation activity of Bat can be applied to find optimized results in VM placement.  To avoid local minima, it generates Non-Allowed Set of VMs inclusion of under-loaded and overloaded VMs which are analyzed for task allocation.   In the centralized system, different metrics and threshold values are maintained to support the number of processes running locally and globally in DC.  In distributed systems, these metrics are maintained in their shared storage.  These metrics are updated periodically by the VM Manager and Task Scheduler running in the local Host Server (HS).  A separate task running in the central system which periodically checks VMs status obtained from local VM Managers which will update number of PMs with under utilization and overloading metrics to select VMs for migration. The proposed Heuristic vs. Meta-heuristic approaches for VMs Allocation and Migration (HM-VM-AM) in cloud computing algorithm considers the Non-Allowed Set generated by the Bat algorithm. The number of VMs belongs to the same PM in this set is compared with that PM’s threshold value.  If it exceeds, it immediately updates the relevant metrics in the centralized system which initiates the migration process immediately, without waiting for that specific migration process’s periodic checking.  HM-VM-AM considers the existing efficient heuristic method named, IRIAL (Improved Resource Intensity Aware Load balancing) for selecting the destination PMs for VMs migration.  It applies an Improved Multi-Criteria Decision Making (IMCDM) method to find lightly loaded PMs as well as heavily loaded PMs with less completion time of all its tasks. It also considers the resource intensities for selecting destination PMs which reduces the number of VMs migration, communication cost and time, and performance degradation.  So the incorporation of job allocation and migration process would reduce the overhead of centralized system and maintains the SLAs.