2005, the grid will be capable of six tera-
flops (trillions of operations per second).
This will eventually rise to 15 teraflops,
equivalent to eight times the power of
the ASCI Red supercomputer at Sandia
National Laboratories. The China grid
enables each university to specialize its
research in one area of computer num-
ber-crunching—climate modeling, say,
or fluid dynamics—while at the same
time giving students access to the
computing resources of all the other
specialist universities. This isn’t just
an efficient way to deploy hardware,
it’s an efficient use of human scientist
“wetware” too.
Fluid dynamics is such a preposter-
ously complicated issue that it seems
to appear on the list of uses for virtu-
ally every supercomputing technology.
Companies like Boeing and the Airbus
consortium are accustomed to building
massive computing clusters to run their
virtual wind-tunnel models; frankly,
they can probably afford to maintain
their own systems like this. But just as
the web has democratized information,
the Grid will bring computing power
to the little man. This means smaller
companies and smaller nations won’t
always need to band together into para-
lyzing consortia in order to tackle major
engineering projects.
We’re still a long way from a Global
Grid—it took the web 10 years to
move from invention to ubiquity—but
grid computing has unstoppable
momentum already. We all know about
Moore’s Law, which posits that the
number of transistors that can be fab-
ricated on a single integrated circuit
doubles every 18 months, but networkcapacity has been doubling more than
twice as fast over the same period.
Grid computing provides a way to har-
ness this power. In the home, we can
expect projects like SETI@home and
Folding@home to be followed by proj-
ects deployed through web services,
without the need to download a sepa-rate client. Next will come integrated
operating system support, possibly
within three years. After that, the desk-
top portion of your operating system
and application programs will begin to
shrink as more of the load is carried by
the emerging Grid.
In the next 10 years, we might
well have transitioned to a world
where home computers are stateless
machines that boot from ROM and
execute all application-level soft-
ware via a global Grid system that
remotely supplies data to our system
faster than the processor can pump
data over the AGP bus today. When
that happens, we will ask ourselves,
“Where did all the data come from
before we had the Grid?” nBEST EFFORT The scheduling algorithm
used by the Internet today. All packets
are (mis)treated with equal priority.GIGE (GIGABIT ETHERNET) An
Ethernet version with a speed of
1Gb/s. The most advanced version of
Ethernet to date is 10GigE.INTERNET2 Trademarked name for a
U.S. university and corporate project
to develop advanced network appli-
cations. The reason why everyone
else has to refer to the “Internet Mark
II” to describe the future Internet,
instead. [Let them try and sue this
article’s title – Ed]KERBEROS A network authentication
mechanism that uses a central secret-
key repository, making it useless for
general Internet use.LATENCY How long it takes a packet to
travel from its source to its destination.
Also called delay, lag, and “I thought
this was supposed to be broadband.”MIDDLEWARE The layer of software
between the network and the applica-
tions, defined as “the intersection of
what network designers and application
developers each do not want to do.”PKI (PUBLIC KEY INFRASTRUCTURE) A
much more secure network-authentica-
tion mechanism than Kerberos, but tragi-
cally too complicated to implement.QOS (QUALITY OF SERVICE) A meth-
odology used by Internet routers to
prioritize data traffic. Time-critical traf-
fic (think voice-over-IP calls, or data
exchanged between a game server and
a client) is sent first, while lower-prior-
ity traffic (such as simple file transfers
and DNS requests) is sent afterward.THE JARGON
The concepts behind National LambdaRail and grid computing,
naturally, come with their own baffling array of technical jar-
gon. Here’s a breakdown of the most important terms to know
as we trot toward the future of the Internet.44 MAXIMUMPC MARCH 2005
SON OF INTERNET
Oh, that’s so much clearer now...
“The essence of grid computing lies in the efficient and
optimal utilization of a wide range of heterogeneous,
loosely coupled resources in an organization tied to
sophisticated workload management capabilities or infor-
mation virtualization.”—MATT HAYNOS, Program Director of Grid Marketing and Strategy
at IBM, explaining the Grid in layman’s terms.