We were thinking about what we provide as NIKHEF for WP4 (and WP6-8) for the
9-month release.

Use case:
---------
we will provide a farm plus infrastructure to run LHCb and Alice jobs via
the Grid. An LHCb job and a Alice job are defined as an executable a.out
with an input file which produces an output file. Typically the experiments
may want to run the job 100 times with different input files each time. To
make the rpoblem more interesting, we assume that Alice will need another
version of Linux than Alice, which in reality may be more than true.

NIKHEF Compute Resources:
-------------------------

A farm of 25(to be verified) nodes hidden behind a network switch. On the
switch is a "farm server" and a "file server" which can "see" the nodes. The
farm server and the file server are also on the NIKHEF network in the
standard way.

There is a "LHCb server" and an "Alice server" on the NIKHEF network. These
machines are in principle under LHCb and Alice control although, in
practice, it may be us for the 9-month prototype. On an LHCb server (and the
same holds for Alice) a.out can be built and run, i.e. the necessairy tools
and librabries are available there (or through there) to do so.

On the farmserver run the Globus tools. We assume a user from LHCb (Eric van
Herwijnen for example) can be identified and get a certificate to run on our
farm. The same holds for user Fedrico Carminati for Alice.

On the farmserver runs the WP1 batch to which LHCb and Alice can submit
jobs.

On the farmserver runs the WP4 service which publishes the NIKHEF resources
to the Grid.

Question: are the farm server and the Datagrid server on which runs the
Datagrid (and Globus) software the same machines? Or is the Datagrid protal
a different function alltogether?

We decided we determine the boot partition for the farm nodes. We may say we
have boot partitions with Linux 6 and Linux 7 and mirically these are the
two versions needed by LHCb and Alice. Apart from that we don't provide
anything. The rest has to come with the job parameters when the job is
submitted into the batch. All the rest should be able to be pulled down from
the LHCb and Alice servers.

Somewhere on the Datagrid portal or on the farmserver there also are the
"policies". A database probably which says how many machines are for LHCb
and Alice and many other policy decissions. These can be set by the NIKHEF
Datagrid resource manager

Kors Bos