Separating computation and storage with storage virtualization

Recent advances of hardware, software, and networks have made the management and security issues increasingly challenging in PC usage. Due to the tight coupling of hardware and software, each one of the hundreds or thousands of PCs connected in a networked environment has to be managed and administrated individually, leading to a high Total Cost of Ownership (TCO). We argue that a centralized storage of software and data, while distributed computation in clients, i.e., transparent computing, can address these challenges potentially and reduce the complexity with reduced software maintenance time, improved system availability, and enhanced security. This paper presents a novel approach, named StoreVirt, to realize transparent computing, which separates computation and storage from inside a single physical machine to different machines with a storage virtualization mechanism. With virtualization, all the OSes, applications, and data of clients are centered on the servers and scheduled on demand to run on different clients in a ‘‘block-streaming” way. Therefore, due to the central storage of OSes and applications, the installation, maintenance, and management are also centralized, leaving the clients light-weighted. Further, due to timely patching and upgrading, the system security can be improved. Experimental and real-world experiences demonstrate that this approach is efficient and feasible for real usages. 2010 Elsevier B.V. All rights reserved.