CAREER: Automated Storage Manager - An Operating System for Input/Output Devices on Storage Area Networks

University Of New Hampshire, Durham NH

Investigators

Abstract

Storage systems have evolved from simple disks under the control of file servers to large, indepen- dent disk-array systems connected to the network using Network-Attached Storage (NAS) and Storage Area Network (SAN) technology. A unique feature of these storage systems is that they are directly accessible by applications running on independent computers without the intervention of file servers. A downside, however, is that these systems are very difficult to manage because of the increased de- mands placed on them by data intensive applications and the number of storage devices connected to the network. We propose to improve the management of such storage systems by automating the storage management tasks and incorporating performance techniques to provide the system with the \intelligence" to adapt to varying storage and workload conditions. In essence, this storage manage- ment software system will control and manage all access to storage devices on the network, much as an operating system manages and controls all access to a computer system. The storage manager will allow applications to express their response time, availability, and bandwidth requirements, and then use this information along with its knowledge of device behaviors and application access patterns to decide how best to map data to devices such that application constraints and storage system goals are met. Based on this work, we anticipate the development of user-friendly, self-managed storage systems that are more cost-effective, efficient, and available than current systems that are configured and managed manually. Our education goal is to integrate performance evaluation tools and techniques into the Computer Science curriculum at the University of New Hampshire. Students who will work on the storage management system will require performance evaluation skills and our education plan addresses this issue. Our long-term goal is to coordinate with industry and other institutions to implement a self-managing storage system and make itavailable for research and educational purposes.