[Beowulf] New HPCC results, and an MX question
joachim at ccrl-nece.de
Wed Jul 20 13:42:03 PDT 2005
Patrick Geoffray wrote:
> Greg Lindahl wrote:
>> I am referring to a comparison of the HPCC "random ring latency" to
>> the HPCC "average ping-pong" on the same hardware, with the same
> The random ring latency will increase with the size of the cluster,
> whereas the average pingpong will not as the pair of nodes are ordered
> and ordered nodes are likely to be in the same crossbar. If you
> randomize the machine list, then there is no difference between the
> random ring latency and the average pingpong.
> On a tiny cluster, all nodes are on the same crossbar, so it does not
> matter if the pair are ordered or not.
All this is true, but the MPI library plays an important role, too. On
our 32node Dual-Athlon Cluster, the HPCC random ring latency with the
latest MPICH-GM (using shared memory for intra-node) was 27,9us, while
our MPI/PC-32 has 16,8us. This is on the exact same hardware, using the
same GM driver (measured Sept. 2004), 32 MPI processes on 16 nodes. And
our MPI is (sort of) a swiss-army knife MPI... I assume that the
intra-node communication makes the difference here.
Joachim Worringen - NEC C&C research lab St.Augustin
fon +49-2241-9252.20 - fax .99 - http://www.ccrl-nece.de
More information about the Beowulf