The Study Of Performance Optimization For I/O Virtualization Based On Petri Nets

Cloud computing as a way to offer the most advanced computing services is applied to more and more companies that use the cloud host to instead of physical servers.Virtualization technology is the core of cloud computing,which provides flexible computing resources services by using resource virtualization methods.Virtualization technology includes the CPU,memory and I/O virtualization.CPU and memory virtualization technologies have become increasingly perfect and the I/O virtualization research has become the one of hot spots in virtualization research.The performance of I/O virtualization determines the utilization of physical equipments and user experiences,which play a main role for the development of cloud computing.Fully virtualization Virtual Machine(VM)can be installed directly.It can run a variety of business operating systems(such as Windows,Mac OS etc.)to satisfy the diverse needs of users.However,the CPU operation models will be switched back and forth between root and non root modes when executing I/O operations.It also increases the switching loads to decrease the I/O equipment performance when loading the context of virtual machine and Virtual Machine Monitor(VMM).Recently,the I/O equipment models and the I/O equipment scheduling algorithms still cannot be resolved completely in the researches of I/O virtualization performance optimal.Moreover,they also do not consider the formal modeling and verification.It will case to a very large security hidden danger in cloud computing platform.In this thesis,an I/O instructions set is given to improve the performance of virtual machine I/O devices.It analyzes the binary code that is generated by I/O instructions set module.Generate a number of I/O instruction set,trigger the execution of VMexit with the units of I/O instructions,and switch to the VMM in which the I/O instructions set executive control module to control the execution of I/O instructions set.Finally,the real data of I/O instructions in VM is changed by the memory sharing modules.VM will reschedule its operations to obtain the final I/O operation data.Petri nets are a formal modeling and verification tool with graph representation.They can be used to describe and analyze concurrent discrete event systems.They can be used to describe the dynamic behavior of systems with static structures.The I/O of the instruction set is used to perform control module.It controls the I/O instruction sets of execution in the VMM and manage the resource permissions of I/O instructions set by using the I/O permission bitmap.To eliminate the I/O instructions set to access I/O resource and update the I/O permission bitmap in the presence of security risks,this thesis uses Petri nets to model the modules.The reachable marking graphs and the S_invariants of Petri nets are used to analyze and verify the model to ensure their correctness,exclusiveness,and synchronization.Refer to the established Petri net model,the performance optimization of I/O virtualization is designed in the open-source Xen virtualization platform by using C language.The experimental results show that the proposed optimal design can significantly reduce the switching frequency of the CPU,reduce the switching overhead,and improve the performance of I/O devices.