UK National Grid Service - Grid Computing: Infrastructure, Service, and Applications - page 148

• WP9 road mapping: To ensure the NGS future development and

deployment strategy is planned and accounts for changing tech-
nologies and user-requested services.
ETF: To provide an interface between the NGS activities and those

•

of the ETF. This ensures that the ETF undertakes work appropri-
ate to that of the NGS and ensures the NGS is aware of technologi-
cal changes.

6.5
The current NGS, based around four cluster-based computer systems and
the UK's High Performance Computing services, entered full production
in September 2004, following a preproduction test phase that started in
March 2004. During the period from March 2004 to June 2006, the NGS
continued to grow in both numbers of users and partner institutions.
The NGS is made up of different physical types of nodes but it can be
useful to classify them as one of two types, compute nodes or data nodes.
Compute nodes are powerful in terms of computation whereas data nodes
provide more storage. The additional storage supports database activities,
through the provision of high-performance Oracle systems, and the ability
to store other large datasets as requested by end users. Each node also has
a useful software environment that enables high-performance parallel
programs to be developed and executed. The base level of software
required to be compatible with the NGS is dei ned in the NGS Minimum
Software Stack (MSS). This document dei nes for prospective resource
providers what middleware software they need to deploy, in order to pro-
vide interfaces that are compatible with the NGS. To ensure compatibility
the NGS also monitors all certii ed sites on an ongoing basis, and regularly
reviews partner sites to ensure they continue to meet their dei ned service-
level dei nition.

Core Founding Member Sites

6.6
In phase 1 of the NGS, the four core sites were split into two types of
resource. The provisioned hardware was provided via a tender exercise
by ClusterVision.
At the University of Oxford and the White Rose Grid, University of
Leeds site, two “compute clusters” were installed. These were designated
as compute clusters as they were provisioned to primarily focus on users

Core Site Resources