Systems

Our Systems

Raad2 is one of our two flagship systems, procured from Cray, Inc. in 2016. It runs the Cray Linux Environment (CLE), an optimized derivative of SuSE Linux Enterprise Server, and it employs SLURM as its workload manager. Cray has since been acquired by HPE hence this is now considered an HPE system. Raad2 is served by a DDN ExaScaler (Lustre) storage system.

Raad2

HPC Vendor

Cray (now HPE)

System Model

XC40

Interconnect Technology

Aries

Interconnect Topology

Dragonfly

Number of Nodes

192

CPU Cores per Node

24

RAM per Node

128 GB

Total CPU Cores

4128 (Intel Xeon Haswell)

Peak Performance

120+ TFLOPS (Linpack)

Storage Model

DDN - ExaScaler 7990X

Parallel Filesystem

Lustre

Storage Network

FDR Infiniband: 56 GB/s
Peak Read Bandwidth: 16 GB/s

Aggregate System Disk

800 TB

Raad2 also has eight V100 Nvidia GPUs housed in four external nodes integrated into the rest of the system via an FDR Infiniband fabric. We usually refer to this (logically) separate system as Raad2-gfx or sometimes simply call it the "GPU cluster" since this is the only GPU system we have at the moment.

Raad2-gfx

GPUs model

NVIDIA V100

Number of GPUs

8

Hazeem is the 2nd of our two flagship systems. It too is a Cray (and now HPE) system acquired in 2018. It began life in the custody of the Qatar Environment & Energy Research Institute (QEERI) and was recently transferred to our group (RCCG) in Q2 2025. Hazeem is nearly identical in architecture to Raad2 and in fact uses the same software stack, although being an XC50 it is one generation newer than Raad2 (which is an XC40). Unlike Raad2, Hazeem is served by a Sonexion Lustre storage system.