From:                     Kendra Smith

Sent:                      Monday, May 22, 2000 8:15 PM

To:                         M?crosöft Research Tech Talk, Sem. Notice

Cc:                         Kendra Smith

Subject:                 UW-CSE Colloq / 5-22-2000 / Yelick / Berkeley / System Support for Data-Intensive Applications

UW-CSE Colloq / 5-22-2000 / Yelick / Berkeley / System Support for Data-Intensive Applications

 

*NOTE* This lecture will NOT be videotaped.

 

UNIVERSITY OF WASHINGTON

Seattle, Washington 98195

 

Department of Computer Science and Engineering

Box 352350

(206) 543-1695

 

COLLOQUIUM

 

SPEAKER:      Kathy Yelick, UC-Berkeley

 

TITLE:          System Support for Data-Intensive Applications

 

DATE:           Monday, May 22, 2000

 

TIME:           4:30 pm

 

PLACE:                   EE1-125

 

HOST:           Susan Eggers

 

ABSTRACT:

 

Computing is headed towards an era in which the focus is shifted away from

desktop machines towards millions of small embedded devices, backed by

large-scale service providers.  The digital information explosion will be

accelerated the vast array of input devices, including digital cameras,

recorders, and sensors, making data storage and retrieval two of the key

services for large-scale providers.  If done well, both the devices and

the services will be invisible parts of our environment; if done poorly,

our lives will be chaotic.  There are many research problems that must be

solved to attain the former goal, including better user interfaces,

increased scalability, more reliable systems, and new kinds of public

policies. 

 

In this talk, I will concentrate on the service side of this "Post

PC" era, and discuss some of the problems that arise in storing,

accessing, and computing on enormous data sets.  One of the most difficult

challenges is the these large-scale systems is the balance between high

reliability and high performance.  I will briefly describe ISTORE, a

reliable, scalable system architecture for storage-intensive

applications.  At the application level, languages like Java avoid

certain kinds of programmer errors, and therefore lead to more reliable

applications, but many programmers are reluctant to use them because of

performance concerns.  I will talk about two projects, Titanium and

Sparsity, which address these performance concerns by using

domain-specific information about the applications to achieve high

performance.  Both of these projects were originally designed for

scientific applications, but we have recently shown that they are

well-suited to information retrieval problems that appear to be critical

in handling the flood of data in the next generation of

computing.  Throughout the talk, I will give examples of some of the

exciting research opportunities that exist in Computer Science and in

areas that overlap with medicine, biology, environmental engineering, and

other interdisciplinary areas.

 

Email: talk-info@cs.washington.edu

Info: http://www.cs.washington.edu