Scientists at CERN announced yesterday that eight major computing
centres have managed to sustain an average continuous data flow of
600 megabytes per second for 10 days. It is a significant milestone
for scientific grid computing.
The total volume of data transmitted between CERN, the
European Organisation for Nuclear Research near Geneva, and seven
sites in the
US
and Europe – amounting to 500
terabytes – would take about 250 years to download using a typical
512-kilobit per second household broadband connection.
In basic terms, grid computing can be described as a network
of computers and data storage systems, brought together to share
computing power. Where a computer is not being used, or is using
only a fraction of its power, the grid will allow that power to be
used by someone else.
The concept differs from the World Wide Web, which only
enables communication through browsers, because it actually allows
access to computer resources. It is also different from
peer-to-peer computing, which enables file-sharing between two
users, because it allows sharing of resources among many, not just
two.
The potential of computer grids is enormous and when the
concept becomes mainstream it holds the promise of transforming the
computer power available to the individual. At present, a computer
user is restricted by the power of his own computer. When the grid
comes on line there will be no restrictions: the cheapest, oldest
model will have access to the computing resources of millions of
other computers worldwide.
Scientists at CERN are collaborating with scientists worldwide
in the creation of what is hoped will be the world's largest
computer grid, in order to analyse the massive volume of data that
will be produced when CERN's latest and largest ever particle
accelerator (known as the Large Hadron Collider, or
LHC
) becomes operational in 2007.
The exercise completed yesterday was the second in a series of
four service challenges designed to ramp up to the level of
computing capacity, reliability and ease of use that will be
required by the worldwide community of over 6000 scientists working
on the LHC experiments.
Other participants included Brookhaven National Laboratory and
Fermi National Accelerator Laboratory (Fermilab) in the US,
Forschungszentrum Karlsruhe in Germany,
CCIN2P3
in
France,
INFN-CNAF
in Italy, SARA/
NIKHEF
in the Netherlands and Rutherford Appleton Laboratory in the
UK
.
"This service challenge is a key step on the way to managing
the torrents of data anticipated from the
LHC
," said
Jamie Shiers, manager of the service challenges at CERN. "When the
LHC
starts operating in 2007, it will be the most
data-intensive physics instrument on the planet, producing more
than 1500 megabytes of data every second for over a decade."
Fermilab Computing Division head Vicky White welcomed the
results of the service challenge.
"High energy physicists have been transmitting large amounts
of data around the world for years," she said. "But this has
usually been in relatively brief bursts and between two sites.
Sustaining such high rates of data for days on end to multiple
sites is a breakthrough, and augurs well for achieving the ultimate
goals of LHC computing."
In fact the test exceeded expectations by sustaining roughly
one-third of the ultimate data rate from the LHC, and reaching peak
rates of over 800 megabytes per second.
The next service challenge, due to start in the summer, will
extend to many other computing centres and aim at a three-month
period of stable operations. That challenge will allow many of the
scientists involved to test their computing models for handling and
analysing the data from the
LHC
experiments.
