DATE: Thursday, March 4, 2004
TIME: Noon - 1 pm
PLACE: Hamerschlag Hall D-210

Larry Huston
Intel Labs, Pittsburgh

Diamond: A Storage Architecture for Early Discard in Interactive Search

As storage capacities continue to increase and digital data is more readily available, it has become increasingly difficult to find specific items in a large collections of complex data (digital images, protein structures, etc.).

This talk describes a system, Diamond, which has been designed to enable interactive applications that search large, un-indexed data sets. Diamond enables efficient search through "early discard"; eliminating the bulk of the irrelevant items as early as possible. Diamond provides additional efficiency by re-ordering search tasks and by dynamically partitioning computation between host computer and the storage devices to adapt to different hardware configurations and network bandwidths. This talk will present some performance measurements using our application for searching digital images.

Larry Huston is a researcher at Intel's research lab in Pittsburgh. Larry previously worked as a software architect for Intel's Network Processor Division. Before joining Intel Larry worked at NetBoost and Ipsilon Networks. He received his Ph.D. in Computer Engineering from the University of Michigan in 1995. He also holds MSE and BSE degrees in Computer Engineering and Aerospace Engineering from the University of Michigan.

Seminar Host:
M. Satyanarayanan

For Further Seminar Info Contact:
or visit