Research On Virtual Machine Scheduling In Real-time Environment

With the development of computer hardware and driven by demand,virtualization technology has become a hot research topic of current computerscience. By using the virtualization technology, we can improve the utilization andthe flexibility in the allocation of hardware resources. In virtual desktop and cloudcomputing environments, the use of virtualization is progressively accommodatingdiverse and unpredictable workloads as being adopted. Unpredictable workloadsresource allocation will cause great difficulties due to the virtual machine monitorlack of understanding internal workload of each virtual machine. If we ignore thework load inside a virtual machine, the virtual machine scheduling algorithm willseriously affect the IO performance. Existing scheduling algorithms lack ofunderstanding of the internal work load. This will seriously affect the IOperformance in the case of a mixed workload, increase the IO response delay, andcan not be applied to real-time environment.In this paper, we present improved methods of credit scheduling algorithm ofXen. By using gray-box knowledge reasoning technology, we are able to identify theIO task in the internal workload of each virtual machine. By using these informationof the internal workload, we propsed a BOOSTing mechanism based on theselection. Selectively raise the priority of the virtual processor to give priority to thecorresponding field of the IO Tasks improve the response speed, and reduce theresponse time to achieve real-time purposes. In addition, we propose fourcomplementary mechanisms for the shortcomings of the proposed BOOSTscheduling algorithm based on the selection. The four complementary mechanismsare the cancellation BOOST priority mechanism, reducing Credit precise mechanism,setting Rate value mechanisms and UNDER queuing mechanisms. These fourmechanisms can effectively fill the insufficient of BOOST scheduling algorithmbased selection, and improve the stability of the algorithm.At the end of the paper, we do the experiment On Xen virtual machine monitorto verify BOOSTing scheduling algorithm based on the selected can inmprove theIO response in mixed workload.The experimental contents include the Credit scheduling algorithm of Xen, theproposed BOOST scheduling algorithm based on the selection and complementarymechanisms. The experimental results show that the proposed algorithm fluctuationscan substantially reduce the response in terms of mixed loads, or pure I/O load, and the average of response time is very small. So the proposed algorithm can be used inreal-time environment.