|
More palm/explore links:
+ Quick Start Guide
+ Overview of System Resources
+ Filesystem Access and Policies
+ Programming Environment
+ Batch Queues
+ Software and Tools
|
 |
Batch Queues on the
SGI Altix 3000
Read the documentation below the
table to familiarize yourself with
the restrictions on some of the
queues below. All queues except
debug and pproc will run on e1,
e2, and e3. The default number
of CPUs for any job is 2, and the
default wall clock time is 5 minutes.
Special Queues on the SGI Altix
3000 System
| |
Up to 1 Hours |
Up to 3 Hours |
From 1 to 12 Hours |
Longer than 12 Hours |
| Up to 16 processors |
- datamove max 2 proc/job and 2 hour maximum
|
|
|
|
| Up to 32 processors |
|
|
|
|
| From 2 to 254 processors |
|
|
|
|
| More than 254 processors |
|
|
|
|
Numbers in parentheses designate
priority, with 7 being the highest
priority and 1 being the lowest.
debug
- Routed specifically
to the system’s
front end, palm
- Time constraint
of 1 hour maximum per
job
- No more than 4 jobs in this queue can be run by the same user at the same time.
datamove
- This queue is to be used for the purpose of moving data (data
archival jobs, staging jobs, no MPI/OpenMP processing).
- Only 2 jobs allowed per user at one time to run.
- Job size is limited to 2 processors.
- There are only 10 processors in total set aside for this queue.
- Jobs in this queue will run on the backend systems (not palm, but rather e1, e2 or e3).
- If users want to use this queue they need to specify the queue name in their qsub parameters "-q datamove" on the command line or "#PBS -q datamove" in the job script itself.
pproc
- This queue is to be used for pre/post processing work (not
production runs).
- Jobs now limited to no more than 16 processors
- If users want to use this queue they need to specify the queue name in their qsub parameters "-q pproc" on the command line or "#PBS -q
pproc" in the job script itself.
- No more than 6 jobs in this queue can be run by the same user at the same time.
general_small
- No more than 5 jobs in this queue can be run by the same user at the same time.
- Job size is limited to 18 processors.
- Minimum wallclock time has been lowered to 1 hour by default (so the queue allows 1-12 hours walltime).
general
- No more than 8 jobs in this queue can be run by the same user at the same time.
- Allows jobs sized between 16 processors and 254 processors.
background
- Designed
to have the lowest priority
on the system and is targeted
for two groups:
- NCCS staff
- Users who have
used up all their allocated
hours on the system
- This queue is only turned
on if there isn't any work waiting in the high_priority, general_hi or
general queues.
- Note
that users may still
run jobs in the background
queue even if they have
a current allocation; use of
this queue will not count
against that allocation
amount.
general_hi
- Designated
for users who have a demonstrated
need for either very large
or very long jobs
- Must be approved
by NCCS staff, the NCCS Director,
or NASA HQ
- Total number of processors
available (and maximum number
of processors per job) in
this queue is 510
high_priority
- Reserved
for use on jobs designated
by NCCS staff, the NCCS Director,
or NASA Headquarters as needing
priority within the overall workload
- Total
number of processors available
(and maximum number of processors
per job) in this queue is 510
|
|
|
