Performance of Commercial Applications on Shared-Memory Multiprocessors

Wednesday, April 21, 1999 - 17:30
TH 331
Kourosh Gharachorloo Western Research Lab Compaq Computer Corporation

The market for large multiprocessor servers has enjoyed a phenomenal growth during the past few years. A decade ago, large multiprocessors were part of a niche market and were primarily used to solve difficult scientific and engineering problems. Today, commercial applications such as databases and Web servers constitute the largest and fastest-growing segment of this market. This evolution is continuing with the increasing popularity of the Web.

Our research at WRL has focused on characterizing the processor and memory system behavior of commercial applications. This talk presents an overview of the studies we have done during the past three years, and describes the performance implications for various processor architectures and memory system designs.


Kourosh Gharachorloo is a research scientist in the Western Research Laboratory at Compaq Computer Corporation (formerly Digital Equipment Corporation). His research interests are parallel computer architecture and software, including hardware and software distributed shared memory systems, and the study of database and webserver applications. Before joining Digital, he was a key contributor in the Stanford Dash and Flash projects. In addition, his thesis research on memory consistency models has influenced academic research and impacted implementations of several commercial microprocessors.

Recently, he has been involved in the development of Shasta (a software DSM system) and in the study of commercial applications such as Oracle and AltaVista. In addition, he has contributed substantially to the design and specification of Alpha processors and server platforms (both 21264- and 21364-based systems). He has also been a contributor to the Alpha architecture specification, and has authored a chapter of the Alpha Architecture Reference Manual.

Gharachorloo received a BS in electrical engineering, a BA in economics, an MS in electrical engineering, and a PhD in electrical engineering and computer science, all from Stanford University. He has authored over 35 technical conference and journal papers, and has filed over 10 patents. He has also taught several short courses at Digital, lectured numerous times at Stanford, and served on various invited panels and program committees.