header

BRAF HPC Resources

 
C-DAC Logo
 

BRAF HPC Resources

PARAM BioEmbryo

PARAM BioEmbryo cluster is a 100 teraflop machine which achieves its performance in 30:70 ( CPU:GPU) ratio. PARAM BioEmbryo is based on AMD EPYC processors with zen2 microarchitecture. It has 16 CPU only nodes and 4 GPU nodes containing 2 GPU cards in each node. Each node is a dual socket node with two 32 core AMD processors, which accumulates to 1280 CPU cores in the cluster. The cluster has 5120 GB RAM distributed across 20 nodes. Total usable storage of 240TB capacity is present as Luster based PFS mounted across the cluster nodes. The cluster is having 100 GBps EDR infiniband as primary interconnect.

It has the open source CentOS operating system and is configured with an OHPC cluster stack that consists, compilers, and libraries. Widely used bioinformatics applications are pre installed for the users.

PARAM Bio MBRYO

Hardware Overview:

Cluster Details
Peak performance 100 TF
Number of Master nodes 2
Number of compute Nodes 20
Node type Dual socket rack based servers
Processor
  • AMD EPYC 7502 CPU
  • Total 1280 compute cores
Total RAM 5120 GB distributed across 20 nodes
Total Storage Luster PFS with 240 TB usable space
Interconnect 100 Gbps EDR Infiniband
Other Rack enclosure with contained cooling and backup
CPU only nodes
No of nodes 16
Node specification Dual socket node with total 64 cores
Processor 2 x AMD EPYC 7502 CPU
RAM per node 256 GB
GPU nodes
No of nodes 4
Node specification Dual socket node with total 64 cores
Processor 2 x AMD EPYC 7502 CPU
RAM per node 256 GB
GPU 2 x NVIDIA Tesla v100 GPUs

Software overview:

Operating System CentOS-7.7.1908
Cluster Suite OpenHPC
Scheduler SLURM
Compilers GCC, AOCC
Libraries ACML
Application software Gromacs, NAMD, BWA, Blast, GATK
PARAM BioEmbryo Software Stack
PARAM BioEmbryo Software Stack

PARAM BioInferno

PARAM BioInferno cluster is the new addition in BRAF which enhances its capability to handle Big Data applications along with traditional HPC applications. PARAM BioInferno is based on the unique design architecture solution which fulfills the need of traditional computing as well as future computing. It consists of heterogeneous hardware with hybrid capabilities of executing HPC as well as Big Data Jobs. It is connected with the 100 GBps HDR Infiniband network as the primary interconnect.

PARAM BioInferno

Hardware Overview:

Cluster Details
Peak performance ~ 150 TF
Number of Nodes 34
Node type
  • CPU only nodes
  • GPU nodes
  • High memory nodes
  • SMP node
  • Vector node
Processor
  • AMD EPYC 7502 CPU
  • Total 2048 cores
Total RAM 25 TB distributed across cluster
Total Storage
  • Luster PFS with 1.5 PB usable capacity
  • 1.5 PB of HBFS across all nodes
Interconnect 100 Gbps HDR Infiniband interconnect
CPU only nodes
No of nodes 24
Node specification Dual socket node with total 64 cores
Processor 2 x AMD EPYC 7502 CPU
RAM per node 512 GB
Local Storage 32 GB HDD
High Memory nodes
No of nodes 4
Node specification Dual socket node with total 64 cores
Processor 2 x AMD EPYC 7502 CPU
RAM per node 1024 GB
Local Storage 32 GB HDD
GPU nodes
No of nodes 4
Node specification Dual socket node with total 64 cores
Processor 2 x AMD EPYC 7502 CPU
RAM per node 512 GB
Local Storage 32 GB HDD
Accelerator 2 x NVIDIA Tesla v100 GPUs
SMP node
No of nodes 1
Node specification 4U chassis with 8 socket node SMP 192 cores
Processor 8 x Intel Xeon Platinum 8260 CPU
RAM per node 6 TB
Vector node
No of nodes 1
Node specification Dual socket node with 24 cores
Processor 2 x Intel Xeon Gold 6226 CPU
Accelerator 8 x NEC 10BE vector cards
RAM per node 196 GB

Software overview:

Operating System CentOS-7.7.1908
Cluster Suite Vendor customized
Scheduler SLURM
Compilers GCC, AOCC
Libraries ACML
Application software Gromacs, NAMD, BWA, Blast, GATK, HBAT