Load Balancing For Internet Applications In Hybrid Cloud

Since online shopping is becoming increasingly popular, ecommerce websites are facing intensive user requests from an number of customers. Especially in promotion seasons, these websites need to address flash crowds which pull heavy pressure to private infrastructure and even render the websites unresponsive. Hence, to maintain fast response time and availability, the e-commerce vendors need to deploy more infrastructure to handle flash crowds of user requests. However, it will cost enormous expense to buy physical servers. Moreover, when the traffic returns to normal, these physical servers are in idle, which is a waste of money.Different from buying physical servers, leasing virtual resources of IaaS clouds which provision elastic, cheap, and theoretically infinite resources, is much more economic. However, trading process always involves much confidential data which e-commerce vendors hesitate to outsource to the public cloud. Hence, a hybrid cloud solution, which contains the private cloud and the public cloud, is promising in dealing with flash crowds. However, the bursty and fluctuating flash crowds make it challenging to distribute user requests with goals of delay-minimizing and cost-saving.In this paper, by adapting Lyapunov optimization techniques, we design an online decision algorithm for request distribution and capacity adjustment so as to achieve the average response time arbitrarily close to the theoretically optimum and control the outsourcing cost based on a given budget. The simulation results demonstrate that in a hybrid cloud, our load balancing algorithm can maitain cost below budget as well as minimize the average response time.