IOT: Pratyush and Mihir - What is Pratush super computer?



IOT: Pratyush and Mihir - What is Pratush super computer?


From the official of Indian Institute of tropical meteorology website introduction, we have:

Pratyush is India’s fastest supercomputer yet (Jan 2018) with 4+ Petaflops of peak performance powered by Intel Xeon Broadwell E5-2695 v4 CPU and has total storage of 27 Petabytes. It has Intel KNL 7210 nodes with a  peak performance of  42+ TFLOPS .




What is Pratush super computer?

From Wikipedia, the free encyclopedia, we have:

Pratyush and Mihir are the supercomputers established at Indian Institute of Tropical Meteorology (IITM), Pune and National Center for Medium Range Weather Forecast (NCMRWF), Noida respectively. As of January 2018, Pratyush and Mihir are the fastest supercomputer in India with a maximum speed of 6.8 PetaFlops at a total cost of INR 438.9 Crore. The system was inaugurated by Dr. Harsh Vardhan, Union Minister for science and technology, on 8 January 2018.The word 'Pratyush' (Hindi: ????????) defines the rising sun.

Being a High Performance Computing (HPC) facility, Pratyush and Mihir consists of several computers that can deliver a peak power of 6.8 PetaFlops. It is the first multi-PetaFlops supercomputer ever built in India.

Pratyush and Mihir are two High Performance Computing (HPC) units. They are located at two government institutes, one being 4.0 PetaFlops unit at IITM, Pune and another 2.8 PetaFlops unit at the National Centre for Medium Range Weather Forecasting (NCMRWF), Noida. Both units and provides a combined output of 6.8 PetaFlops.

Pratyush and Mihir are used in the fields of weather forecasting and climate monitoring in India. It helps the country to make better forecasts in terms of Monsoon, fishing, air quality, extreme events like Tsunami, cyclones, earthquakes, lightning and other natural calamities such as floods, droughts etc. India is the fourth country in the world to have a High Performance Computing facility dedicated for weather and climate research after Japan, the United States and the United Kingdom.

System Configuration Details




Pratush super computer - System Configuration




Hardware




Cray XC40 System                            System Configuration

Cray XC40 Compute Cabinets           18

Interconnect Network                        Cray Aries with Dragonfly topology

Total Peak Performance                     4,006 TFLOPS

Compute Node




Number of Nodes                                          3,315
Processor                                                          Intel Xeon Broadwell E5-2695 v4 CPU (18 core, 2.1 GHz)

Memory Per Node                                           128 GiB Memory DDR4-2400 w/ Chipkill TM technology

Total Memory                                                   414 TiB

Accelerator Node




Number of Nodes                            16
Accelerator                                       Intel KNL 7210, self-hosted mode, single socket per node

Memory Per Node                          96 GiB DDR4-2133 w/ Chipkill TM technology

Total Peak Performance               42.56 TFLOPS

Additional Nodes




External Login Nodes               5
Utility Servers                            8

Utility Racks                               1

Storage System Configuration




Parallel File System                                                             Lustre
IOPS per File System  Home and Scratch                     50,000+

Lnet Nodes                                                                          39

Storage Array (For Home and Scratch)                           Sonexion 3000

Storage Racks                                                                     5 racks + 6 Tape library frames

PFS – Home




Useable Storage                                        8.6 PB @ 264 GiB/s Write 231 GiB/s Read
Sonexion Config                                       One MMU + 26 SSU

Lustre I/O Nodes (Embedded)               2 MDS + 52 OSS (Active/Active config)

HDD Data                                                  6 TB  7.2K RPM on GridRAID (RAID 6)

HDD Metadata                                         900 GB  10K RPM on RAID 10

Number of Files per MMU                      4B+

PFS – Scratch




Useable Storage                                         1.1 PB @ 50.69 GiB/s Write 53.21 GiB/s Read
Sonexion Config                                         One MMU + 4 SSU

Lustre I/O Nodes (Embedded)                 2 MDS + 8 OSS (Active/Active config)

HDD Data                                                    4 TB  10K RPM on GridRAID (RAID 6)

HDD Metadata                                           900 GB  10K RPM on RAID 10

Number of Files per MMU                        4B+

HSM / Archive Tier




Archive Useable                                              27.4 PB
Storage Array Disk Cache for HSM              NetApp E5600 series (273 TB)

HSM Servers                                                    2 Metadata controllers + 2 Connection Managers+                                                                                           4 connector agents +2 External gateways

Tape Library                                                     SpectraLogic TFinity

Number of drives LTO-7                                 48

Media LTO-7 Qty/Slots                                    5190

HSM Solution                                                    Versity VSM 1.0

Software Stack




Cray Linux Environment                                                        6.x
Cray Programming Environment (CPE)                               10 Seats

Intel Parallel Studio XE Professional Edition                       7 Seats

Workload Manager                                                                 PBS Pro 13.0

Additional/Optional debugger                                              2048 core

PV-View License                                                                       1

Previous Post Next Post