|Target usage: large memory, threaded and MPI jobs|
|System information: see bramble system page in web portal|
|System status: see bramble status page|
|Real time system data: see Ganglia monitoring page|
|Full list of SHARCNET systems|
Bramble is an SGI Altix 450 consisting of 32 dual core Itanium 2 (IA-64) processors for a total of 64 core and 128 GB of memory.
Since this is contributed hardware, regular user jobs can be suspended on demand if needed.
Any problems with bramble should be reported via the SHARCNET web portal problem ticket system.
Accessing the System
The machine can be reached via ssh on the SHARCNET network as bramble or remotely as bramble.sharcnet.ca.
Compiling and Submitting Jobs
The default compiler is Intel and will be found in the user's path.
Serial, threaded, and mpi jobs have a 7 day runtime limit while the test queue has a 1 hour limit. There is a maximum of 60 processors available to run user jobs.
Bramble is a NUMA architecture; processes will see faster memory performance in their local memory domain. To ensure that processes always try to allocate memory on their local domain we recommend wrapping your executable with numactl -l in your sqsub command. Presently this behavior is implemented by default for jobs that run in the MPI queue. To run jobs without our default binding policy use -f nobind when submitting your job with sqsub. For further details on using numactl please see Using numactl.