The first phase of the Interactive and Batch Farm is now in place and
available for data reduction and other analysis and simulation jobs
related to the current production experiments in the Experimental
Halls. The Interactive Farm provides an interactive UNIX environment
where codes can be prepared for eventual batch processing on the
central Batch CPU Farm. This first phase of the Batch Farm includes
four dual-CPU Sun 2170 systems (farms2/3/4/5) and four dual-CPU IBM
RS6000/240 systems (farma1/2/3/4). The Interactive Farm currently
includes an additional 2-CPU Sun (ifarms1) and 2-CPU IBM system
(ifarma1) identical to those in the Batch Farm, plus 6 Hewlett
Packard systems (ifarmh1/2/3/4--C160s and ifarmh5/6--9000/735s). The
recent addition of a Linux system to the Interactive Farm
(ifarml1 -- "L" for Linux) will allow Users to port their codes to
this platform, in anticipation of the very likely addition of a Linux
component to the Batch Farm in early Fiscal Year 1998 (Fall, 1997).
All of these systems are fully integrated into the Jefferson Lab
Common UNIX Environment (CUE) (See http://www.jlab.org/ccc/
cue/cue.html) and can access the Jefferson Lab On-Line Batch System
(JOBS) tape and batch utilities -- jget, jput, and jsub (See
http://www.jlab.org/ccc/farm).
In order to provide group-based accounting of the CPU resources
available to experiments, access to all of the farm systems (batch
and interactive) will be managed by "UNIX groups" which contain the
usernames of Users involved with the on-line processing of experiment
data. The software coordinators for Hall A (Bob Michaels, rom@jlab.org)
and Hall C (Steve Wood, saw@jlab.org) have assisted the Jefferson Lab
Computer Center in developing the appropriate UNIX user group structure
to manage access to these dedicated CPU resources for their Halls.
Hall B (Dieter Cords, cords@jlab.org) is currently evaluating the
appropriate group structure for CLAS processing. Beginning Monday,
September 1, 1997, access to the interactive and batch farm systems
will be managed based on these groups. At the current time, sufficient
CPU resources are available to permit all of the experiment groups to
have full access to both the Interactive and Batch Farms. In the
future, this group structure would be used to allocate resources
appropriately between the various groups in different stages of
their processing. By early September, Users will be able to view
system loads and group usage via web-based accounting reports that
will assist users and experiment groups in selecting their processing
platform as well as to monitor the resources their group is using.
ALL Users involved in the off-line processing of experimental data
should verify NOW that they are included in the appropriate UNIX
group(s) for the experiments in which they participate. See
http://www.jlab.org/ccc/groups/. This web page is only accessible
from jlab.org (cebaf.gov) systems (this includes personal computers on
the site), so you must first log into a Jefferson Lab CUE system
(jlabs1, jlaba1, jlabh1, etc.) to run your web browser if you need
to access this information from your home institution. This web
page includes all of the groups and members of those groups currently
identified by Halls A and C. The UNIX groupings for CLAS will
be announced in the near future. If you find that you need to be
moved to a different group, please contact the software coordinator
for your Hall (see above). Please refer other questions to Rita Chambers
(chambers@jlab.org), the Jefferson Lab Computer Center Manager or Ian
Bird (igb@jlab.org), the Computer Center's Physics Computing Manager.
The systems for physics computing will continue to expand and evolve. In
addition to the web-based utilities to provide up to date information on
system loads and usage, plans are underway to expand the Batch Farm with
a Linux component and to expand the RAID-based "Work" disk pool areas.
The most current information can always be found on the CUE home page,
http://www.jlab.org/ccc/cue.html.
Rita Chambers