Enabling Grids for e-Science: The EGEE Project - Grid Computing: Infrastructure, Service, and Applications - page 59

PRODUCTION Cumulative normalized CPU time by REGION and VO (excluded dteam and ops VOs)
EGEE VOs. February 2007-January 2008

1.6e+8

1.4e+8

1.2e+8

Asia Pacific
Central Europe
CERN
France
Germany/Switzerland
Italy
Northern Europe

1.0e+8

8.0e+7

6.0e+7

Russia

South Eastern Europe
South Western Europe
UKI

4.0e+7

2.0e+7

0

FIGURE 3.2 Accounting example.

3.2.2 The Preproduction Service

Before new middleware distributions are deployed on the production
service, they are thoroughly Tested on a small-scale system, encompass-
ing some 20 sites, called the preproduction service. This testing is vital for
such a large grid, as any errors or faults in deployed software would not
only expose many sites to security risks, but would also be time-consuming
to i x owing to the diverse properties and geographical spread of the grid
sites connected to the EGEE infrastructure. Also, the preproduction sites
are in most cases identical to ROCs and thus the preproduction service
allows them to gain experiences with new distributions they can then
pass on to their resource centers.
The preproduction service works in a truly distributed manner that as
far as possible mimics the properties of the production service and allows
the project to test new software in the most rigorous way possible prior to
deployment across the whole production service. A set of predei ned tests
is typically run and users can also test new middleware versions against
their applications.

3.2.3 Certification and Testing Testbeds

Prior to their deployment on the preproduction service, the middleware
components need to be individually tested and certii ed. EGEE operates