From Documentation
Jump to: navigation, search
requin
Hostname: requin.sharcnet.ca
Target usage: Large, tightly-coupled MPI jobs
System information: see requin system page in web portal
System status: see requin status page
Real time system data: see Ganglia monitoring page
Full list of SHARCNET systems


System Overview

Spec Info Remarks
Cores (CPU) AMD Opteron 2.6GHz
Cores/node 2
Memory/node 8 GB 4GB (login)
Interconnect Quadrics Elan4
Storage 70 TB
OS HP Linux XC 3.2.1
Max. Jobs 10000

For system notice/history, please visit the Requin system page in the SHARCNET web portal.

System Access and User Environment

Login Nodes

Requin login nodes have the standard settings, including limits on memory, number of logins and cpu time.

Storage

Each node contains 160 GB of local storage (/tmp) and there is 70 TB of space available as Requin-local /work and /scratch.

Submitting Jobs

Requin currently uses an older scheduler infrastructure, which does not treat job memory use as a schedulable resource. Fortunately, nodes have only 2 CPUs, and the scheduler never shares a node between two parallel jobs, so this is rarely of concern.

Otherwise, sqsub and related tools behave as normal.

Note that requin is the only cluster that features a pre-emptive test queue, such that if you submit your jobs with sqsub -t ... they will start almost immediately. Note that users may only submit 1 job to this queue at a time, and that jobs running in this queue are limited to a duration of 1 hour.