PPPL Cluster Hardware
The PPPL cluster consists of several sub-clusters typically named after
researchers from PPPL who had a powerful impact on fusion energy research.
dawson - the main cluster.
This group of systems, linked via a 10Gbps low latency Ethernet network,
provides the majority of PPPL's scientific computing resources. Each system
contains AMD Opteron CPUs, and 2GB of memory per CPU core.
ellis - a cluster of systems reserved for small (1 to 4) CPU jobs. These
systems contain AMD Opteron CPUs and 2 GB of memory per CPU core.
kruskal - a low latency
cluster of 32 systems, each with 32 CPU cores and 64 GB of memory,
linked via an Infiniband QDR interconnect.
ganesh - these are large memory Symetrical MultiProcessing (SMP)
systems with more than 4GB of memoryi per core.
There are also some smaller clusters constructed for specific
uses and applications, like the mccune and fielder clusters used
exclusively by TRANSP jobs, the greene cluster used by the XGC and
M3D groups, and the gque GPU servers containing Nvidia TESLA GPUs.