National Centre for Medium Range Weather Forecasting: HPC BHASKARA

National Centre for Medium Range Weather Forecasting: HPC BHASKARA

Question: Earth System Science Organisation National Centre for Medium Range Weather Forecasting/ESSO-NCMRWF has recently experienced a boost through High Performance Computing Device/HPC BHASKARA. Give the background of ESSO-NCMRWF and discuss the importance of the HPC.

Background

• HPC facility of NCMRWF in 1988 as a mission mode generated medium range weather forecast and provides agro-metrological advisory services

• Customised forecasts for other applications have been developed for numerous user community as well as different governmental sections

• NCMRWF has also collaborated in providing forecast products for RIMES/Regional Integrated Multi-Hazard Early Warning System countries

• It is also the BIMSTEC/Bay of Bengal Initiative for Multi-Sectoral Technical and Economic Cooperation Centre for Weather and Climate.

• Accuracy of weather prediction is assured through cutting edge technology and understanding of physical and dynamical processes

• NCMRWF focuses on efficient and innovative data assimilation from Indian satellites such as Meghatropiques, INSAT-3D, SARAL/AltiKa etc. and global satellites

• NCMRWF is implemented a coupled ocean atmosphere model for extension of forest cast times to monthly and fortnightly scale

• R&D efforts improve the skill of weather prediction through HPC facility

BHASKARA: Highlights

• HPC facility installed at NCMRWF comprises an IBM iDataPlex Supercomputer of peak computing power of 350 TeraFlops with 67 TeraBytes of aggregate memory

• This HPC facility has been named BHASKARA

• The overall ESSO HPC facility has peak computing power of ~1.2 PetaFlops

• BHASKARA also deploys open source and propriety applications needed for execution, post processing and visualisation of data and applications

• It is equipped with Energy Aware Scheduling Algorithm for optimising energy efficiency for execution of applications
• BHASKARA will be operational from June 2015

• It provides high resolution 10 days global weather forecasts and predicts extreme weather events

• Login and compute nodes of BHASKARA are populated with 2 Intel Sandy Bridge 8 core processors

• This HPC uses FDI 10 Infiniband interconnect in a Fat Tree Configuration as its high speed network for input/output operations and MPI messages

• BHASKARA employs IBM’s General Parallel File system and each compute node is equipped with two 8 core processors of 16 cores with own Red Hat Enterprise Linux OS sharing 64GB of memory with no user accessible swap space

• This HPC is rated at 350 peak TFLOPS with maximal efficiency of 83% across 15 compute racks

• It is connected to 3 PetaBytes formatted high speed disk storage

• Storage is connected to achieve a performance of 20 GB per second per PetaByte of storage

• Software stack is designed to be very flexible keeping the requirements of end users from varied backgrounds in mind

• High end post processing and visualisation are also deployed as part of this HPC facility

Facts and Stats

• The HPC will enable the NCMRWF to provide very high resolution 10 days deterministic weather and probabilistic forecasts within a time window of 5 hours from standard observation time

• R&D for extreme weather events will be facilitated through BHASKARA such as tropical cyclone, heavy rainfall events, cloud burst events and so on needing very high resolution regional models with horizontal resolution of 1.5km or lower

• Enhanced computing power will also enable NCMRWF to generate 50 member ensemble prediction from fully coupled ocean atmosphere modelling system

R&D in data assimilation and modelling for both atmosphere and ocean will be possible through enhanced computing capacity