Special Features:
Interview: CHIEN DISCUSSES SMARR'S
OptIPuter
By Neil Alger, GRIDtoday
In September of 2002, the OptIPuter project recieved $13.5 million dollars
to be awarded over 5 years to push the distributed computing effort forward.
Initially the vision of Cal-(IT)2 director Larry Smarr, the effort has drawn
Grid computing visionaries from accross the field, including USC's Carl
Kesselman, UIC's Tom DeFanti, UCSD's Mark Ellisman and SDSC's Phil
Papadopolous.
Still, for all the high-profile involvement of leaders within the Grid
community, the OptIPuter seems to have flown nearly under the radar for the
last year. GRIDtoday correspondent Neil Alger spoke recently with Dr. Andrew
Chien, who has been called the "chief software architect" of the project, in
an attempt to shed some light on the OptIPuter's progress since September of
last year.
GRIDtoday: To begin, could you possibly elaborate on the nature of your
work
on the OptIPuter project?
Dr. Andrew Chien: The OptIPuter project includes several different threads
of
activity -- driving applications, experimental network testbed design and
construction, and middleware. Larry Smarr is the overall PI for the project,
with other leaders such as Phil Papadopoulos (experimental networks) and Tom
DeFanti of UIC (experimental networks and visualization), and Mark Ellisman
(applications). First, my research group at UCSD is one of the key teams on
the middleware research activity, and in addition I've been called the "chief
software architect" or other terms which connote the coordination of the
larger middleware team which includes Jason Leigh of UIC, Carl Kesselman of
USC-ISI, and many other distinguished and talented researchers.
GRIDtoday: With so many different groups working in conjunction on the
OptIPuter project, it can be difficult to get a sense of the progress that is
being made. Can you give us a brief overview of what has happened since last
September's announcement of NSF funding for the OptIPuter, as well as
elaborate on what sort of timeline and roadmap have been established?
Chien: This is hard to do tersely, but the progress has been significant.
We
have established initial testbeds and plotted a clear experimental network
infrastructure strategy and plan for the next several years. This plan has
already been put into action at the lead site (UCSD) and also at UIC. The
middleware team has developed an initial software architecture and researchers
are working to develop innovative new technologies which can be plugged into
that framework.
GRIDtoday: How is the OptIPuter different from other distributed computing
projects that are currently in progress?
Chien: While there are many innovative things about OptIPuter, I think the
thing that stands out the most is the close coupling of challenging massive
data applications, a critical mass of top-flight middleware researchers, and a
focus on dedicated lambdas in an experimental network environment which
enables innovation across system abstraction layers and cross-fertilization
across groups. While there are many distributed computing projects, I know of
none that are as focused on the remarkable opportunity of dedicated end-to-end
optical connections and which couple the development of middleware with
driving applications to ensure we develop relevant technologies.
GRIDtoday: Because the OptIPuter project is physically based out of UCSD,
it
is situated within a key locus of the Grid community, sharing a campus with
both SDSC and NPACI. What kind of cross-fertilization is there between the
OptIPuter project and these other related groups, and has that proximity
affected the ultimate vision of the OptIPuter in any way? Further, how do you
see the OptIPuter project situated within the Grid community as a whole?
Chien: There is tremendous synergy and cross-fertilization, not only
because
of physical proximity, but also because many of the key leaders of the
OptIPuter project are key players in SDSC and NPACI. While I can't speak for
those efforts, I have little doubt that our vision for OptIPuter will become
an integral part the long-term vision for cyberinfrastructure. It is important
to note that OptIPuter is different from those efforts in that it is primarily
a research project, while those others mentioned are primarily infrastructure
efforts.
GRIDtoday: Whose products are currently 'under the hood' of the OptIPuter?
Do
you foresee an increase in enterprise involvement in the project as it moves
forward?
Chien: We are using "bleeding edge" hardware and software technologies from
many outstanding companies. As the project advances and the scale of
experimental networks and experiments grow, we expect even greater interest
and involvement with leading technology companies.
GRIDtoday: What are the major hurdles that need to be overcome for the
OptIPuter to reach its projected potential?
Chien: There are a long list of hurdles, but let me point out two. First,
one
major hurdle is to establish a simple model of use or abstraction for the
underlying end-to-end optical connections. Without it, few applications will
benefit and such services will not be widely available. With it, virtually all
applications could benefit dramatically. Second, a key hurdle is to gain
acceptance of the notion that applications can express network configuration
control and the network can respond to it in real time. From the network
management side this means dealing with dynamic configuration demands, and
from the application side, this implies reasoning about requirements -- both
when they can and can't be met.
GRIDtoday: Is there anything else that you would like to add for our
readers?
Chien: The OptIPuter is a large project with a large number of outstanding
researchers. I've doubtless failed to describe all of the exciting key
challenges and technologies, but I hope that I've piqued your interest enough
that you'll keep an eye on what we're doing!
About Dr. Andrew A. Chien
Andrew A. Chien is the Science Applications International Corporation Chair
Professor in the Department of Computer Science and Engineering at the
University of California at San Diego. Andrew Chien leads the Concurrent
Systems Architecture Group and is involved with joint projects with both NCSA
and NPACI. Furthermore, he is a co-founder and CTO of Entropia Inc, an
enterprise desktop Grid company.
|
