Research On Resource Scheduling Algorithms In Cloud Computing Environment

Cloud computing as a new service provision model distributes user’s tasks to different data centers so that applications can access computing power, storage space and information services as needed from cloud data centers. Cloud data centers leverage virtualization technology to abstract the hardware and software resources forming pools of virtualized resources. Through resource scheduling techniques those resources were provided to users with "on demand, pay as you go" style. With the rapid increase of users and data center size, how to deploy such resources quickly, efficiently, and dynamically has become an important issue for cloud resource scheduling. Thus, how to efficiently use data center resource while ensuring that the quality of service does not violate the service level agreements(SLA) is a major problem for cloud resource scheduling.Cloud resource scheduling is facing two key issues:balancing the system load and minimizing the operating cost of data centers. To address the first issue, we propose a modified simulated annealing load balancing (SALB) resource allocation algorithm for cloud computing. It sets the objective to minimize the standard deviation of the hosts’ load and balance the entire system’s load. Different from the traditional SA that selects the initial solution and produces the neighborhood solution randomly, we select the appropriate initial solution and produce a new neighborhood solution based on real-time system load conditions. Through the VM migration technology, it migrates the VMs running on a high utilization host to a low utilization host which can accommodate them. Meanwhile SALB accepts a suboptimal migration with a certain probability toward reaching the global optimal solution. Then we realized the proposed resource scheduling algorithm SALB on an expanded cloud computing simulation platform CloudSim. The experimental results show that SALB can achieve better load balance of cloud systems, and performs better than the traditional simulated annealing and round-robin algorithm.To address the second issue of minimizing the operation cost of data centers, we propose a simulated annealing combined genetic algorithm (SACGA) as a virtual machine resource allocation strategy. By adding the simulated annealing ideas to the crossover and mutation process in the traditional GA, the GA accepts a suboptimal solution with a certain probability to avoid the premature phenomenon of falling into a local optimal solution. Simulation results show that SACGA is able to achieve better data center operation cost than the basic genetic algorithm. Finally, we conclude the thesis and explain some future work.