Eos cluster downtime June 13-14
The Eos cluster will be unavailable from 9am on Monday, June 13, to 5pm on Tuesday, June 14, for expanding the cluster.
The expansion downtime will involve adding three Infiniband (IB) switches to the IB fabric to accomodate the new nodes and performing needed firmware and software upgradres.
PLEASE NOTE: We are proceeding with the cluster expansion with only a subset of the IB cables needed for full bisectional bandwidth between the current IB switch and the new IB switches. We expect the remaining IB cables to arrive and be integrated at the end of the month.
Forty-eight additional nodes will be added for this expansion. Each node will have 12 cores per node, using two Intel Westmere 6-core processors, rather than the 8 cores per node on our existing compute nodes. Four of these nodes will also have NVIDIA GPUs based on the Fermi architecture. The total number of cores for Eos will increase from 2592 cores to 3168 cores after the expansion is completed.
Posted on: 2:24 PM, June 9, 2011