 |
|
|
|
|
|
|
|
|
|
|
NCCS Discover Cluster
General Architecture
- The system has been designed to be very scalable, starting with a base unit and quickly adding additional scalable compute units. The base unit will allow contain the basic file system components and initial compute capability. This will allow the NCCS to integrate the base unit into the center with minimal disruption to the existing HP system. The base unit will have the following capabilities:
- 3.3 TF Peak
- 128 Nodes/256 Sockets (all compute)
- 3.2 GHz Intel dual core Dempsey processors
- GPFS & 60.0 TB DDN (raw)
- SilverStorm Infiniband with full bisection bandwidth between all nodes in the base unit
- The system is designed so that subsequent scalable units can be integrated into the cluster environment with minimal to no disruption. Users of the base unit will be able to take immediate advantage of the compute capability as additional units are added.
- The subsequent scalable compute units will be approximately twice the size in terms of the number of nodes and compute capability than the base unit. While the exact configuration of the scalable compute units is still begin architected, in general the units will be approximately:
- ~6 TF Peak
- 256 Nodes/512 Sockets (all compute)
- Dual core processors
- GPFS & ~90 TB DDN (raw)
|
|
|
|
|
|
|
|
