SGE

From Jacobson Lab Wiki

Jump to: navigation, search

Our queueing system is SGE v6.1u3.
The full user manual does not appear to be available anymore at [1]

Queues may be specified by name with qsub, for example: "qsub -q broadway queuefile.sge"
Our headnode is timesquare.jacobsonlab.org. Jobs for all queues must be submitted from this host.
We have a variety of queues:

broadway:

    20 processors, 2core Xeon nodes, primarily for Loose jobs, 32bit. [aging]


timesquare:

    2core Opteron nodes, For mixed use of MPI, MD, Loose jobs. [aging]


schrodinger_masters:

    35 processors ?? nodes, For loose jobs. [aging]


8core:

    80 processors. 8core Xeon nodes, primarily for MPI MD Simulations and docking, 64bit.


ib:

    128 processors. For mixed use of plop, (MPI) MD, docking, loose jobs, etc. [newer]


gpu:

    3 gpu machines, 32 cores each (31 CPUs and 1 GPU)[I believe]. Primarily for long MD (GPUs and CPUs). Architecture is not compatible with latest plop. [newest]
    To use the GPUs on this machine, you submit to a separate queue now: cuda. 


cuda:

    The GPUs. Only for long MD.
Personal tools