Research On Memory Resource Cooperation And Sharing In Cloud Computing Platform Based On Virtualization Technology

With the rapidly expanding Internet information service and increasing users reliance, Internet service providers face the challenges of delivering better service at lower cost. While cloud computing gradually becomes their choice, for the reason of cloud computing's pay-as-you-go model and its'super elevation cost performance. Based on virtualization technology, cloud computing center can deliver at-scale and on-demand services in an abstract fashion from underlying components, with traits of easier manageability, less maintenance and high utilization.However the processing of virtualizing computing resource holds two challenges. The first is performance; the second is the semantic gap between the virtual machine and the service. Besides, as one of three most critical computing resources, memory becomes considered the bottleneck, as it is harder to be multiplexed and shared, comparing to CPU and I/O. In response to these difficulties, this paper makes some research on memory resource cooperation and sharing in cloud computing platform based on virtualization technology. The main contributions of this thesis are as follows:1. Based on analysis of the Para virtualization platform Xen, this paper proposes and implements a method to dynamically collect Virtual Machines'(VMs) memory usage information at Virtual Machine Monitor(VMM) side, by using improved balloon driver installed in VMs. In this way, we can bridge the semantic gap between VM and VMM.2. Based on the analysis of the existing proposals, VM's free memory division view and VM's memory division view are proposed to enhance memory efficiency and achieve service level agreement. Corresponding memory scheduling algorithm is designed and implemented to provide a hierarchical memory service to VMs co-located in single physical host. This mechanism makes VMs' memory resource elastic and which is a base of on-demand service.3. This paper has found and remedied some faults in Xen's Content-Based Page Sharing mechanism and made it practicable. Experiment results show that our work can reduce memory pages redundancy, increase memory utilization, and improve system I/O performance.4. Using the technique mentioned above, this paper designed some mechanism to support commercial application. Users can get multi-scale memory choices and pay-as-you-go service.To sum up, our research is a beneficial exploration of memory resource cooperation and sharing technology in system virtualization. It has good theoretical and practical value in constructing cloud computing platform which has magnanimity data processing capacity. Supported by the National Natural Science Foundation of China and the National Basic Research Program of China, our works has actually brought dramatic benefits to commercial cloud computing center.