Green, more efficient computing

[ Silence ] 6>>One of the things that’s
happened in the last 5 to 10 years has been that
computers have got cheaper and cheaper, that supercomputers or high performance computers
are now actually made from large collections
of commodity pieces. And that has enabled
manufacturers to place thousands if not tens of thousands of
chips into very small packages or footprints, as
they call them. If you’ve ever sat
with your laptop on your lap watching
your television at night, then you know how warm
that device can get. Now, if you can imagine a
thousand machines running like that, and probably
actually appreciably hotter than that chip, you’ll get some
sense of the kind of energy draw and cooling that’s required
to make these machines run. [ Silence ]>>There’s no free
ride in computing, so anytime you do any
kind of a transaction, an eBay transaction,
you do a Web search, you look at your local newspaper
online, you send some e-mail, somewhere that’s
drawing some electricity, and not just the
electricity that’s being used to power your machine, but electricity that’s
being drawn elsewhere to provide those
services for you. [ Silence ]>>All the large
providers of services now, whether it be Amazon or eBay or
Google or Yahoo or Microsoft, any large institution like a
university, hospital systems, government agencies, are all
running a large data center, some of them very large. And these large data centers
aren’t like the data center that we maybe think of
when we think of them. We think of the data center
of the HAL 9000 in the 2001: A Space Odyssey, or in other
movies from the ’70s or ’80s with large banks
of white cabinets, very quiet environment,
a cool environment. Modern data centers
are hot places. They’re noisy places. They’re like a boiler room. [ Silence ]>>I think we’re
going down two paths. First of all is how can we use
the equipment we have right now at its most effective, and secondly are there
new technologies emerging that will allow us to make
a significant improvement in our efficiency? On that second front we’re
definitely very actively exploring right now with
some brand new companies, some very interesting
opportunities in high performance computing,
that will require some work on our side to make
these machines effective. But do allow us the opportunity to consider significant
reductions of energy usage. So one reasonable measure, there
would be a reduction of 75% or 80% of the cost
of electricity and the associated cooling, requiring some different
coding skills and some different
analysis skills, but that opportunity
is absolutely there, and we want to see if
we can leverage that. [ Silence ]>>Another one of the little
known but widely observed facts of computing use is that most
people don’t use their computer most of the time. Most of the time the
machine sits idle. And that’s true in large
data centers as well. For example, you got a
machine that runs payroll, how many times are you
going to run payroll? You’re only going to
run it when you need it. And when you’re not running it, then the machine
basically sits idle. We have a fairly aggressive, in fact we have a very
aggressive program here at Purdue that cycle scrapes,
or captures waste cycles or machines just sitting idle. So whether they be in our labs or whether they be our
large research machines, when they’re not actually in use
with a user in front of them, they’re actually being
used by other programs to carry out research. My own desktop machine,
the 2 Processor Dell, also is one of the
7,000 machines on campus that we use for this purpose.

Daniel Ostrander

Related Posts

Leave a Reply

Your email address will not be published. Required fields are marked *