A total of 67Gb is currently available for central scratch space. This space is shared amongst all users. Each user has a directory within the central scratch space allocated to them. This directory is automounted as
/scratch/bohrg/<username>on both storage and dominion1 (the head/submit nodes), and on all compute nodes.
All compute nodes have scratch space available to users on a local/internal disk. I/O to local scratch space should be faster than to central scratch space.
The amount of local scratch space available varies from node to node, but a minimum of 8.2Gb should be available on all nodes:
...
...
On some machines hosted by Manchester Computing scratch space is regularly "tidied up", e.g., files created more that a month ago are automatically deleted. At present (2005 Apr 19) there is no such automated deletion. Should scratch space become short this automation might be introduced.
There are currently no scratch space quotas in operation. If space becomes short quotas will be introduced.
Scratch space is not backed up. Do not store any files in scratch space which you need to keep or cannot easily reproduce (e.g., from a re-run of a job).
...previous | up (conts) | next... |