Efficient Disk Scheduling For Virtual Machines Based On I/O Performance Isolation

In cloud computing environment, virtualization technology has become an importantmanagement tool. At the same time, virtualization is an indispensable key technology ofbuilding a cloud infrastructure. The ever growing number of I/O intensive VMs sharingdisk device brings the disk I/O resource contention, because the hypervisor cannotprovides I/O performance isolation among VMs. This contention leads to unpredictableI/O completion times, thus affecting service of quality (QoS) of VMs and greatlyinfluencing the cloud user experience in the cloud virtualization environment. Therefore,disk resource allocations are required to enable I/O performance isolation among VMs.Performance isolation has long been a challenging problem for disk resourcesallocation in virtualized environments. While there have been many research works inaddressing the I/O performance isolation and disk utilization, none of works has addressedthem as a whole. To this end, based on investigating the impact of current disk I/Operformance isolation schemes on disk I/O utilization, the studies report that currentisolation schemes bring the unnecessary disk idle and reduce the overall disk I/Operformance due to ignoring the disk states and characteristics of requests. Accordingly, anadaptive proportional-share I/O scheduling framework which exploits the DyIso (dynamicisolation) algorithm is proposed in hypervisor, named iShare. While iShare ensures I/Operformance isolation through proportionally allocating time slices according to theweights of VMs, it also preserves high disk efficiency by detecting disk states andadaptively adjusting the time slice size based on characteristics of requests.Through a series of typical tests in Xen virtualization environment, iShare is provedits effectiveness. The experimental results show that iShare ensures I/O performanceisolation while improving disk I/O efficiency: compared with Blkio (i.e, the default I/Operformance isolation method in Xen), iShare increases disk I/O bandwidth by50%andslightly improves the I/O performance isolation for the sequential write applications. SoiShare can be used as a highly efficient I/O resources allocation solution in the cloudvirtualization environment.