India is poised to launch 18 new petaFLOP supercomputers dedicated to weather forecasting, a move aimed at enhancing the precision and resolution of climatic predictions. These advanced machines will considerably upgrade forecasting capacities, specifically at the block level, providing more accurate cyclone predictions with increased lead time, and detailed ocean state forecasts.
Understanding FLOPs in Computing
FLOPs, an abbreviation for Floating-Point Operations per Second, are utilised as a metric for assessing computational performance and efficiency in high-performance computing (HPC) and artificial intelligence (AI). This includes mathematical computations involving real numbers with fractional parts. With the help of floating-point encoding, extremely long numbers can be handled with relative ease.
FLOPs, however, should not be seen as the sole measure of a computer’s performance. Other factors such as memory bandwidth, latency, and architectural features also play a significant role. Nevertheless, FLOPs do serve as a basic benchmark for comparing computational capabilities – predominantly in tasks that are heavily dependent on floating-point calculations.
The Units of Computing Speed
Computing speed is measured in three main units: Teraflops, Petaflops, and Exaflops. A Teraflop signifies a unit of computing speed equal to one million million (1 trillion) (10^12) FLOPS. A Petaflop, on the other hand, is indicative of a computing speed equal to 1000 TFLOPS (10^15). Lastly, an Exaflop represents a unit of computing speed equivalent to one billion billion (10^18) FLOPS.
India’s Existing PetaFLOP Supercomputers and Their Usage
Currently, The National Centre for Medium-Range Weather Forecasting (NCMRWF) houses ‘Mihir,’ a 2.8 petaFLOP supercomputer, while the Indian Institute of Tropical Meteorology (IITM) is home to ‘Pratyush,’ a 4.0 petaFLOP supercomputer. However, these existing supercomputers, which were commissioned in 2018, will be phased out once the new petaFLOP supercomputers are operational.
As per the agreed terms, NCMRWF will be allocated eight PFLOPs of computing power, with the remaining ten PFLOPs being designated to IITM. This distribution is specifically designed to meet their unique weather forecasting needs. Notably, India’s initial encounter with supercomputing was in 1991 with the launch of the PARAM 8000.
Global and National Supercomputing Rankings
On a global scale, the world’s most powerful computer in terms of PFLOPs is Hewlett Packard Enterprise Frontier, also known as OLCF-5, which boasts peak performance capabilities of 1,685.65 PFLOPs. Meanwhile, within India, Airawat PSAI is recognized as the nation’s largest and fastest AI supercomputing system, clocking an impressive speed of 13,170 teraflops.
Overall, this recent development signals a significant leap for India in the arena of high-performance supercomputing, with a specific emphasis on enhancing weather forecasting capabilities.
Source: IE
