Research On Resource Allocation And Task Scheduling For Green Cloud Computing

With the rapid development of cloud computing, the servers’ scale of cloud data center isconstantly expanding every year, which causes huge power consumption. Furthermore unreasonablescheduling policies lead to energy waste, making the cloud data center operating costs continuelyincrease. To address the issue of high energy consumption and low efficiency of cloud computing,this thesis is conducted a series of research and exploration, mainly includes:Firstly, a highly efficient and reliable green cloud computing system model is designed. Thisthesis focuses on resource allocation module and task scheduling module. Aimed at the actual needsof the two modules, implementation scheme is given. Resource allocation module makes systemresource allocation more reasonable, improves the utilization rate of system resources and reducesthe overall energy consumption of cloud computing system. And task scheduling module is toensure a reasonable task scheduling mechanism for cloud computing system.Secondly, abstract the task scheduling problem to the virtual machine deployment issue with thevirtualization technology. The demand of resources for the next period of task requests is calculatedby combining with a conservative control strategy. Then, a multi-objective constrained optimizationmodel of power consumption is established, and a low-energy resource allocation algorithm basedon probabilistic matching is proposed. Based on this algorithm, another low-energy resourceallocation algorithm is designed with the improved simulated annealing algorithm, in order toreduce power consumption further. Experiment results show that the prediction and conservativecontrol strategy make resource pre-allocation catch up with demands, while improving real-timeresponse ratio and stability of the system. Two algorithms can both activate fewer hosts, achievebetter load balance among the set of high applicable hosts and maximum utilization of resources,and greatly reduce power consumption of cloud computing systems.Finally, a method called multi-level load assessment method for cloud computing taskscheduling is proposed for green cloud computing task scheduling. On this basis, a novel strategy ispresented, called adaptive task scheduling strategy based on dynamic workload adjustment. Withthis strategy, task nodes can adapt to the changes of load at runtime, and obtain tasks in accordancewith the computing ability of each node that realizes self-regulation, while avoiding the complexityof algorithm, which is the prime reason making the master node be the system performancebottleneck. Experiments show that the strategy is a highly efficient and reliable algorithm, which makes the heterogeneous cloud clusters more stable, faster and load balancing. Furthermore, itsperformance is superior to current task scheduling strategy.