Energy aware load balancing for computational cloud

Abstract

Cloud computing is novel technology, which enables any resource as service on demand. Cloud environment motivates highly dynamic resource provisioning. Hence clients can scale up or scale down their requirements as per their demand. Load balancing is very important and complex problem in cloud environment, because of its heterogeneity of the computing nodes. In order to realize the full potential of cloud computing it is vital to minimize energy consumption along with effective load balancing. The aim of Energy Aware Load Balancing (EALB) model is to minimize energy consumption with load balancing. EALB model classifies the incoming job request either CPU bound or I/O bound according to their purpose and behaviour. This classification details are maintained in a table named Pattern History Table (PHT) and organized as hash table. One of the virtual machine (VM) is selected dynamically based on best fit allocation policy and the job is assigned to the victimized VM. From the pattern history table job's nature is identified. Using Dynamic Voltage Frequency Scaling (DVFS) scheme the selected VM's processor clock frequency is increased if it is found CPU bound else decreased (I/O bound). Thus, EALB algorithm saves considerable amount of energy and proves to be more efficient.