Perspective Study on Load Balancing Paradigms in Cloud Computing

Abstract

Due to the growth of IT industry, the need of computing and storage has increased manifolds. Cloud Computing is the emerging technology for online allotment of computing resources and storage for user’s data on the pay-as-you-go basis following utility computing model. Cloud computing is a general term for the delivery of hosted services over the Internet and includes virtualization, networking, utility computing, distributed computing, software and web services. Considering the growing importance of cloud, it must provide high performance gain to the user and must be beneficial for the Cloud Service Provider at the same time. With this goal, there are challenges ahead. Load balancing being one of them which helps meeting the QoSbenchmark that the user requires and on the other hand maximizes the provider’s profit by optimum use of resources.In order to balance the load in cloud the resources and workloads need to be scheduled in an efficient manner. Various scheduling algorithms are used by load balancers to determine which backend server to send a request to. It is that selected server which allocates resources and schedules the job dynamically on some virtual machine (VM) located on the same physical machine. Service provider is responsible to dynamically reallocate or migrate the VM across physical machines for workload consolidation and to avoid over-utilization or under-utilization of resources. Matching QoS requirements with a cost-effective amount of resources is challenging as workloads, user demands take large swings over time. Prediction is necessary as the virtual resources that cloud uses have their non-negligible setup time. A proactive dynamic provisioning of resources, that estimate the future need of applications in terms of resources and allocate them in advance, releasing them once they are not required, will be helpful.