cclab
cluster
The cclab cluster consists of six machines, each with
dual-core/dual processor Intel Woodcrest Xeon 5130
processors (that's effectively four processors per
machine). These machines are shared with the Jones Group
and are also used for teaching. The master node is
cclab-m0. It has 16 GB RAM and two mirrored 500 GB disks.
The five compute nodes are cclab-w1 through w5. They have 4
GB RAM and 80 GB disks.
Job policy
Jones Group users have priority on w1, w2, and two
processors of w5; Stern group users have priority on w3,
w4, and the other two processors of w5. Node m0 is to be
reserved for jobs that require a large amount of RAM, and
jobs run by users taking courses.
If users run a job on the master node or on a compute node
for which they do not have priority, they should start
their job with 'nice +19 ...' (or use 'renice +19' if the
job has already started). Please note that it is OK to run
a job on any node as long as there are processors available
(it is not OK if there are already 4 jobs running).
However, you should run with nice and be prepared to
terminate your job on a node for which you don't have
priority. One or two extra niced jobs will not adversely
affect the performance of high-priority jobs, but any more
than that will.
Disk usage on cclab
Jobs running on cclab
Automatic backups do not exist at this time, so users are
responsible for backing up their own data.