Research

Elizabeth Varki
Department of Computer Science
University of New Hampshire
Kingsbury Hall
Durham, NH 03824

Phone: (603) 862-2319
Fax: (603) 862-3493
Email: varki@cs.unh.edu
www: www.cs.unh.edu/~varki

My research interests

Research Summary


Storage systems represent a growing market; in recent years there has been an explosion of applications (which include scientific ``grand-challenge'' programs, multi-media systems, and large transaction-based information systems) with varying performance needs that use enormous amounts of data. These applications have high Quality of Service (QoS) requirements from storage devices, irrespective of the location of the data and its users and the problems that could interfere with data access. It is very difficult to coordinate the storage, network, and computation resources required for these heterogeneous applications. A solution to this problem of storage data management is to have the storage system manage its data. This approach was first proposed by Gelb [Gelb89] who referred to it as system-managed storage. Attribute-managed storage [Borowsky98], a formalization of the system-managed approach, is currently being studied in Hewlett-Packard Storage Labs. In addition, companies like EMC, Veritas, and IBM and research labs like the NASD Lab in CMU are investigating the design and development of intelligent storage systems.

Storage technology has moved from a system-bus interconnect to a network interconnect with the development of Network-Attached Storage (NAS) and Storage Area Network (SAN) technologies which allow storage devices to be directly connected to the network. However, storage management, an increasingly difficult job, still remains under the control of administrators and operating systems. My work proposes the development of an independent storage management system that will control all access to network storage devices, much like an operating system which controls all access to a computer system. This storage management system will support multiple application types, server platforms, and operating systems with the help of an address-mapper that converts all virtual addresses generated by clients to actual storage addresses. In addition, the system will allow applications to express their response time, availability, and bandwidth requirements, and will use this information along with its knowledge of device behavior and application access patterns to decide how best to map data to devices such that application constraints and storage system goals are met. This platform independent storage management system will allow the development of large network storage systems that are more cost-effective, efficient, and available than current systems which are configured and managed manually.

Current grants

Students

Collaborators

Publications:

Helpful Internet Sites:

Return to my Home Page