CDAC shall also setup a Linux based Open Source PARAM Supercomputing Cluster consisting of 5-nodes; 1-master and 4-compute nodes with low latency Infiniband interconnect. This facility also includes 100TB of storage. Interconnects are low latency high-speed network which include Infiniband, Gigabit and Ethernet switches along with essential software. One 42U racks with required number of power sockets, Cables and Connectors. This facility will be deployed at NUST, Windhoek, Namibia.
HPC Cluster Specifications:
- Total Rpeak: 17.7 TF (Teraflops) (Pure CPU 10.7 TF+ GPU 7 TF)
- Total Nodes: 4 CPU only nodes + 1 GPU node
- Total Cores: 5320 (CPU 160 cores + GPU cuda cores 5160)
- Total Memory: 480 GB
- Storage: 100 TB + 54 TB Backup
Primary Communication: Dell H1024-OPF 24-Port
Secondary Communication: Dell N1524 24-Port
IPMI Communication: Dell N1524 24-Port
Tape Library: PowerVault TL1000
The facility will be used for running various applications from different domain areas like Bioinformatics, Climatology, Astrophysics, Computational Fluid Dynamics, Computational Chemistry, Molecular Dynamics, Finite Element Analysis, Agriculture etc., CDAC will also provide demonstration in Applications from Climatology and Bioinformatics domain.
This Facility will give boost to current research on High performance computing by adopting next generation Hybrid architecture. It will help the users and researchers to prepare their code for future architecture.
It will also comprise of CDAC's indigenous products InClus, Onama and CHReME:
- InClus is a web-based tool for managing multiple Linux systems from a central location. It has many useful features for general system administration.
- ONAMA is a GUI based Installer and Execution Model that supports installation and execution of a well-selected set of parallel and serial applications across several engineering disciplines such as Computer Science, Mechanical, Electronics & Communication, Electrical and Chemical engineering etc. Besides, it consists of a number of CUDA (Compute Unified Device Architecture) enabled applications.
- CHReME (CDAC HPC Resource Management Engine) is a web-based portal that empowers scientists, researchers, system administrators with intuitive GUI to exploit the resources of HPC systems. CHReME provides GUI for creation, submission, monitoring and management of jobs. It allows users to configure their execution environment through compilers and libraries selection, scheduling parameters etc. Also, the engine facilitates administrators to perform various management tasks such as scheduler configuration, hosts management, compilers and libraries settings etc. Besides, it contains a workflow based WRF portal, which provides in-depth GUI for configuration of WRF scientific parameters and WRF data pre-processing, main execution and post-processing.