|
Features: Feb. 18, 2002
ISSAC Will Analyze Large-Scale IT Systems
As large-scale IT systems, such as the World
Wide Web, become increasingly complicated, scientists and
engineers are working on new ways to organize them. A team
of Rensselaer researchers has developed "ISSAC,"
the Intelligent System for Speculative and Active Code,
which will allow bulky programs and networks to run faster
and more efficiently with fewer problems.
| |
A team of Rensselaer researchers has
developed "ISSAC," the Intelligent System
for Speculative and Active Code, which will allow bulky
programs and networks to run faster and more efficiently
with fewer problems.
|
|
A key feature of ISSAC's "brain"
is PerfMiner—the performance-mining engine whose intelligence
will analyze the problems of large-scale, data-intensive
distributed applications.
ISSAC will start by probing all the different
parts of a program and record how the pieces work together.
PerfMiner will then analyze the data, suggest solutions
for performance problems, and ultimately make the program
run more smoothly.
"Unlike other performance analysis
systems, PerfMiner uses sophisticated data-mining techniques
to automatically infer the causes," says Mohammed Zaki,
assistant professor of computer science. "Then it works
in a continuous learning framework to find better solutions."
The research is supported by a three-year
National Science Foundation grant through its Next-Generation
Software Program. Computer scientists Christopher Carothers
and Bolek Szymanski are also leaders of the research team.
|
