2005-2006 Academic Year

Speaker: MARK PULLEN
CS SEMINAR
GRID COMPUTING OVERVIEW


Availability of inexpensive high-performance computing clusters
(tens to hundreds of processors at 1GHZ or higher clock), coinciding
with availability of affordable high-performance wide area networks
(hundreds to thousands of Mb/s), has spawned interest in a new
form of distributed computing called "grid computing". This new
approach is a logical extension of cluster computing to the WAN,
where clusters in multiple locations are shared under a common
software such as the Globus Toolkit. Many institutions, GMU
CS department included, now have clusters and also have access to the
Internet2 and the new National LambdaRail optical network. It
is logical to consider coupling these clusters into a larger 
resource,
however there are significant issues in the areas of scheduling/
resource allocation and other barriers such as firewalls.

Being interested in grid computing for distributed, highly reliable
systems, I attended the SURA Grid Computing workshop in January 2005.
While it is clear this is still a "bleeding edge" area, it also is
true that grid computing shows much promise and should be on the
GMU CS planning horizon. I will review what I learned at that
meeting in this seminar. Attendees are encouraged to add to my
presentation with more facts and discussion.