Modifying Broker Policy for Better Distribution of the Load Over Geo-distributed Datacenters
School of Information Science and Engineering/East China University of Science and Technology
摘要：As an increasing number of businesses move toward Cloud based services, issues such as reduce response time, optimize cost, and load balance over data centers are important factor that need to be studied. Selecting the suitable data center to handle the user request is affecting those factors directly. The Broker policy determines which data center should service the request from each user base; so choosing appropriate policy can improve the performance noticeably. One of the benchmarks policies is service proximity-based that routing the request to the data center, which has lowest network latency or minimum transmission delay from a user base. If there are more than one data centers in a region in close proximity, then one of the data centers is selected at random to service the incoming request. However, other factors such as cost, workload, number of virtual machines, processing time etc., are not taken into consideration. Randomly selected data center gives undesirable results in terms of response time, data processing time, cost, and other parameters. this work propose modifying that policy by applying new schedule algorithm that control the load balance. the results showed that the using of this algorithm instead of the random selection would improve the distribution of the workload over the available datacenters noticeably.