[RndTbl] threaded CP exists

Grigory Shamov Grigory.Shamov at umanitoba.ca
Mon Feb 23 09:28:06 CST 2015


About 6 days it takes, using 2 nodes with 6 cp processes on each of them,
using MPI-parallel version of cp and copying from one shared FS (Lustre)
to another (NFS). Threaded version of cp (i.e. on one node, several
threads) was also tried but the speedup was less than when copying form
the two clients.

Grigory Shamov

Westgrid/ComputeCanada Site Lead
University of Manitoba
E2-588 EITC Building,
(204) 474-9625


On 15-02-21 12:47 AM, "Trevor Cordes" <trevor at tecnopolis.ca> wrote:

>On 2015-02-11 Grigory Shamov wrote:
>> Hi All,
>> There exist a threaded and (even MPI-parallel) 'cp' for large/parallel
>> filesystems.
>> http://people.nas.nasa.gov/~kolano/projects/mutil.html
>What did it get your cp time down to from 90 days?
>Roundtable mailing list
>Roundtable at muug.mb.ca

More information about the Roundtable mailing list