MSV-2035 Astronomy Document - Inside Design - FINAL - FINAL

Astronomy & Astrophysics 94 5.7.5 Virtual Observatories Virtual Observatories (VOs) are repositories of archival data equipped with not only the option of downloading data but also for analysing data. This builds on the convention in astronomy that all data collected by large facilities are made public after a certain lock-in period to ensure maximum possible usage of data. This has become particularly important in the era of multi-wavelength and multi-messenger astronomy where data from very diverse sources is used together to derive inferences about properties of sources. The capability of analysing data prior to download is essential as some of the raw data products are too large to be downloaded or processed on small facilities. Therefore, in the next two decades, we need to support addition of high end computing along with VO servers to optimise the usage of data archives. Similar support is also required for other dedicated data archives like AstroSat, GMRT, Aditya-L1 and ground-based optical telescopes. As one is envisaging a very large data base, it is desirable to implement various advanced techniques likeArtificial Intelligence andMachine Learning for search and analysis of the data. 5.7.6 Common Facilities forComputationalAstrophysics Computational Astrophysics comprises a variety of applications, from N-Body simulations, hydrodynamic simulations, MHD simulations to detailed analysis of large data sets using statistical techniques as well as modern approaches like Machine Learning. Considerable computational work is also required to check feasibility of planned experiments and facilities and to fine-tune their capabilities to ensure that it will be possible to discriminate between competing scenarios. At present, we have expertise in these areas spread across a large number of institutes and universities. Each institute and university, depending on resources available, provides computing facilities for such work. However, we do not have any functioning equivalent of the national science supercomputing center in the US that provides facilities and support to all researchers in the country. Such facilities provide some amount of computing time and resources free of cost to enable researchers to test and demonstrate capabilities of their codes. Researchers who require more resources need to pay for the same through research grants, or submit proposals for fully-funded computer time. It is worthwhile exploring the provision of a common supercomputing facility for the astronomy community, to be upgraded from time to time. In principle, the National SupercomputingMission (NSM) can provide some of the resources required here. However, in practice, the access to NSM facilities for users beyond the host institute needs to be streamlined with adequate help for customization to the local development environment. It is important to note here that the astrophysics community in India has been making use of machine learning and related techniques for data analysis for several decades. The community has been a leader in terms of code development for research. Therefore, any investment in facilities is likely to result in significant returns in terms of research as well as human resource development. The astronomy community can also explore the use of cloud computing and data storage on the cloud as a potential platform for HPC. This can potentially free up the overheads for purchase and maintenance of hardware and the need to continuously upgrade systems. MEGA SCIENCE VISION-2035