Optimum Resource Allocation of Database in Cloud Computing

Cloud computing is a new generation of computing based on virtualization technology. An important application on the cloud is the Database Management Systems (DBMSs). The work in this paper concerns about the Virtual Design Advisor (VDA). The VDA is considered a solution for the problem of optimizing the performance of DBMS instances running on virtual machines that share a common physical machine pool. It needs to calibrate the tuning parameters of the DBMS’s query optimizer in order to operate in a what-if mode to accurately and quickly estimate the cost of database workloads running in virtual machines with varying resource allocation. The calibration process in the VDA had been done manually. This manual calibration process is considered a complex, time-consuming task because each time a DBMS has to run on a different server infrastructure or to replace with another on the same server, the calibration process potentially has to be repeated. According to the work in this paper, an Automatic Calibration Tool (ACT) has been introduced to automate the calibration process. Also, a Greedy Particle Swarm Optimization (GPSO) search algorithm has been proposed and implemented in the VDA instead of the existed greedy algorithm to prevent the local optimum states from trapping the search process from reaching global optima. The main function of this algorithm is to minimize the estimated cost and enhance the VMs configurations. The ACT tool and the GPSO search algorithm have been implemented and evaluated using TPC-H benchmark queries against PostgreSQL instances hosted in Virtual Machines (VMs) on the Xen virtualization environment