| With growing computing resources and application requirements, there are manychallenges for virtualized resource organization and allocation. Firstly, a package ofresource organization mechanism that performs transparent deployment of service onvirtualization infrastructure is needed. Secondly, the expanding scale of computiingresources put more stress on communication cost of local resource allocation. At last, theperformance of allocation by workload migration could be seriously undermined whentotal resources strain appears to the host. Towards those problems, this dissertation studiesapproached of reosurce organization and allocation in virtualizationg environment.Based on resource requirement of application, service oriented building of virtualizedresource environment organizes computing resources and software resources, and thendeploies service’s running environments with virtual machines automatically. Byintroducing template managing mechanisms and Copy-on-Write based disk image, manyinstances of service are deployed on computing nodes, and so achieved fastbuilding-on-demand of massive service computing environments..Message based allocation of virtualized resource is driven by resource requests. Eachvirtual machine detects its resources usage and send request. The resource allocationmanager re-calculates the virtual machine’s resource quota based on allocation strategy.Strategies to allocate processor and memory resources have been implemented in theallocation framework. The proposed approach can dynamically share the computingresources across virtual machines without too much extra cost introduced to the system.Live migration optimization adjusts the VCPU working frequency according topre-copy speed, so that the downtime of the migration can reach a desired small amount.Comparing to the original live migration, our experiments show that the optimized schemecan reduce up to88%of application downtime with an acceptable overhead. |
PDF Full Text Request