35th Annual Report (2022-2023) - ENGLISH

92 FACILITIES AT IUCAA ComputingFacility IUCAA HPC facility and its scientific impact The IUCAA Computing Facility offers state of the art computing hardware and technology rich environment for IUCAA members, associates and visitors. It also extends an array of specialized High-Performance Computing (HPC) environments to the academic community for their research. HPC serves as one of the most important backbones for modern Astronomy. It is necessary at every stage, for making observations, processing the data, connect the observations to science, making scientific predictions, designing instruments, and planning future theoretical, observational and instrumentation studies. The HPC facilities at IUCAA have been utilised for all these activities. These facilities have led to several major research achievements across different areas in Astronomy. Over the last decade several large scale astrophysical simulations, on various topics such as astrophysical turbulence, large scale structure formation, accretion on to compact objects, AGN feedback and galaxy formation have been carried out at the IUCAA HPC, which have been published in international journals. These simulations require substantial use of HPC systems with good interconnect between compute nodes. IUCAA HPC has also been used extensively to analyse very large scale data, of the order of several TeraBytes to PetaBytes, from observational surveys. Such large data volumes also generate associated data products during the analysis, further expanding the size of data structures. An efficient high performance system with access to large scale fast storage, as set up in IUCAA, is essential for suchworks. A significant fraction of these resources were used for gravitational wave (GW) data analysis. GW signals are faint and hidden in the noise. One needs to search for hundreds of thousands of modelled signals for every second of data in years worth of data, which requires enormous amount of computation power. When the same astrophysical signal is detected in two or more detectors almost at the same time, one can claim a detection. Though the LIGO-India detector is not built yet, researchers at IUCAA and many other Indian institutions are part of the international collaboration with full access to data from the other sensitive detectors in the world, namely the LIGO detectors in USA and the Virgo detector in Italy. The Sarathi cluster at IUCAA is part of the shared resources used by the International Gravitational Wave Network (IGWN) and was used by the national and international users to search for signals in the latest observation runs of these detectors which detected several mergers of compact binary stars, some of whichwere interesting enough to create headlines. IUCAA currently has three major independent HPC clusters dedicated to different applications, namely Pegasus, SARATHI and VROOM. The Pegasus Cluster is to serve the general computing requirement of the astronomy community associated with IUCAA. It has 80 compute nodes, 4 gpu nodes with 32 cores and 384 GB (on old) & 512GB RAM (on new). It uses InfiniBand EDR (100Gbps) as an inter-connect, and Portable Batch System (PBS) as a job scheduler. For visualisation purposes, there are two dedicated graphics nodes equipped with NVIDIA Tesla P100 GPU cards. The cluster consists of more than 2600 Physical cores. The cluster is attached to a 2 PiB parallel file system (Lustre), which is capable of delivering 15 Gbps throughput. Theoretical computing speed of the Pegasus Cluster is 150 TF. The Pegasus cluster has been utilized by about 70 high volume users from IUCAA and various Indian Universities, running applications for Molecular Scattering, Molecular Dynamics, Stellar Dynamics, Gravitational N-Body Simulations, Cosmic Microwave Background Evolution, Fluid Mechanics, Magnetohydrodynamics, Plasma Physics, and the analysis of diverse astronomical data. PresentHPCfacilities The Sarathi Cluster is primarily used for gravitational wave research and is mostly used by national and international members of the LIGO Scientific Collaboration (LSC), which includes many IUCAA members and Associates. The cluster is comprised of heterogeneous compute servers, it is built in three phases. The cluster consists of more than 8000 Physical cores. The theoretical peak performance of the compute node CPUs of the cluster is nearly 530 TFlops. The cluster has 2PiB PFS storage with 30Gbps write and read (1:1) throughput. The Vroom cluster is used solely for the MeerKAT Absorption Line Survey (MALS). This cluster has 21 compute nodes (DELL), 2 MDS nodes, 4 GPU notes and 2 head node which delivers 25 TF computing speed and has a parallel file system (DDN) of 3.5 PiB usable capacity attached to it. The cluster is also attached to 1 PiB archival storage for archiving/serving the processed data to international community. Sarathi Cluster Phase III, Pegasus Cluster, and Sarathi Cluster Phase II are listed at 33th, 45th and 48th rank respectively in the list of top Supercomputers in India published on January 31, 2023. The list is maintained and supported by CDAC's Terascale Supercomputing Facility (CTSF), CDAC, Bangalore. The list is available at https://topsc.cdacb.in/filterdetailstry?pag e=50&slug=January2023 While the HPC facilities at IUCAA have been highly successful in terms of their scientific outcome and high fraction of utilisation of the resources, the facility is due for a major upgrade. Last major upgrade happened about three years ago. A substantial quantum of funding is necessary immediately to maintain the activities and to meet the increasing demands for computation. HPC clusters listed in Top Supercomputers in India Urgentneedforupgrades