Grid Computing

1
Shipping High-Energy Physics data, be it
simulated or measured, required strong national
and trans-Atlantic networking. Within the
Netherlands SURFnet, the Adacemic and Research
Internet Service Provider, boasts a 10 Gbit/s
national backbone. SURFnet interconnects with
Géant at 2.5 Gbit/s, and run several links to New
York and Chicagos STARtap. SURFnet also
strongly supports network research. In their
effort for all-optical net-works, they provide a
2.5 Gbit/s link or lambda to STARLight (Chicago).
This link brings us close to major US
Laboratories like FermiLab and is an important
piece in the research of the EU DataTAG project.
It also provides excellent facilities for the
experiments running at FNAL.
Grid Computing at NIKHEF
NIKHEF collaborates with various universities and
research labs, both in the Netherlands and
abroad. Since 1991 we have operated Condor
Clusters. Grid research was started in 1999 as
part of the WTCW Virtual Laboratory Project, in
collaboration with the University of Amsterdam,
SARA and AMOLF. Current projects include LHC
Computing Grid, DataGrid, DataTAG, CrossGrid and
Virtual Lab for E-Science (VL-E). DutchGrid is
the platform for Large-Scale Distributed
Computing in the Netherlands. It include KNMI,
UvA, Free University, SARA, CWI, AMOLF, CMBI,
LUMC, leiden, TU Delft, UU, and others.
Data shipped over our link to
www.nikhef.nl/grid
www.dutchgrid.nl
Rev. 20020719-2, David Groep, NIKHEF,
ltgrid.support_at_nikhef.nlgt
2
More than a decade ago, NIKHEF was one of the
first five sites in the world to be on the Web.
Developed by Tim Berners-Lee at CERN, Geneva, the
web was meant to ease communications within the
large High-Energy Physics collaboration emerging
at that time. Nowadays, NIKHEF is one of the five
first DataGrid centres. Were the web was limited
to sharing information, the Grid goes further
sharing computational resources, data storage and
large databases without bothering the users with
the how and when. Once plugged-in to the Grid,
all resources in the world are instantly
available to the user. The Grid is widely
regarded as the most viable solution to the LHC
Computing Challenge. When the Large Hadron
Collider (LHC) becomes operational in five years,
a unprecedented amount of data not only has to be
stored, but also analysed. The estimated amount
of data is a few petabyte (1015 bytes) every
year. We estimate that around 50 000 computers
will be needed to process these data. No single
computer centre in the world will be able to
handle the amount of data produced by LHC.
Therefore, the EU DataGrid project has embarked
on a research and development project to solve
the computational and storage problems on the
Grid.
As of March 2002, a functional test bed for Grid
computing has been deployed around Europe. This
test bed was extremely successful for the first
Review of the DataGrid project by the EU
Commissioners. The current application test bed
spans five institutions CERN (Geneva), NIKHEF
(Amsterdam) , RAL (UK), IN2P3 (Lyon) and CNAF
(Bologna). For this test bed, NIKHEF offered a
full suite of services data storage, gatekeepers
to accept jobs, a user interface machine for our
users and a network monitoring system. The
back-end farm consists of 22 worker CPUs, with
more machines on order. In total NIKHEF can avail
over more than 500 CPUs and 200 TByte of
permanent storage.
The Grid Application Cluster in use for DØ
Monte-Carlo production and for external use by
the EU DataGrid (EDG) test bed. This farm is
fully installed using EDG fabric management
tools adding a brand-new system is as easy as
pressing the power button. A new system, complete
with all application software, is up and running
in 15 minutes.
EU DataGrid sites participating in Test Bed 1