What is ECDF? What is Eddie? What is Edikt?
ECDF - Edinburgh Compute and Data Facility
Eddie - the ECDF compute cluster. This is an Intel processor based cluster running 64-bit Linux (Scientific Linux 4).
Edikt - eScience Data Information and Knowledge Transformation project - this project is providing funding and resources for some of the ECDF staff in the startup period.
How do I submit a job?
Documentation is available http://www.edikt.org.uk/edikt2/ECDF_Documentation_Development here.
Will there be any scheduled maintenance of the ECDF "service" that will interrupt service availability?
All planned disruption will be announced through the ecdf-alerts mailing list. See this page for information on how to subscribe.
Is there a job length limit?
Yes - the limit is as little as 30 minutes. The maximum is 48 hours.
If you require more than 30 minutes to run your job, you MUST specify the runtime required when you submit the job.
More information can be found here.
Is there a memory limit?
Yes - the limit for a slot is 2GB of memory. Using a parallel environment will offer 2GB per slot requested. If you require more than 2GB of memory for a single threaded job, use the parallel environment "memory", and request one job slot for each 2GB of RAM you require, up to a maximum of 4 slots.
Can individual users have locally allocated cpu hours?
Yes. There are tools on the cluster for a research group manager to assign quotas to members of their group. More details here
My big parallel job won't run because the cluster is too busy - what can I do?
Use resource reservation - just put a -R y option into your qsub command.
Is there local scratch space available on the worker nodes?
Yes - the directory can be referenced by the $TMPDIR variable. For example, this command would output something into a file in this folder:
myapplication > $TMPDIR/outputfile