Similarity and Location Aware Scalable Deduplication System for Virtual Machine Storage Systems

Cloud computing is widely considered as potentially the next dominant technology in IT industry. It offers basic system maintenance and scalable source management with Virtual Machines (VMs). As a essential technology of cloud computing, VM has been a searing research issue in recent years. The high overhead of virtualization has been well address by hardware expansion in CPU industry, and by software realization improvement in hypervisors themselves. However, the high order on VM image storage remains a difficult problem. Existing systems have made efforts to decrease VM image storage consumption by means of deduplication inside a storage area network system. Nevertheless, storage area network cannot assure the increasing demand of large-scale VM hosting for cloud computing because of its cost limitation. In this project, we propose SILO, improved deduplication file system that has been particularly designed for major VM deployment. Its design provide fast VM deployment with similarity and locality based fingerprint index for data transfer and low storage consumption by means of deduplication on VM images. And implement heart beat protocol in Meta Data Server (MDS) to recover the data from data server. It also provides a comprehensive set of storage features including backup server for VM images, on-demand attractive through a network, and caching through local disks by copy-on-read techniques. Experiments show that SILO features execute well and introduce minor performance overhead.