Pratyush

Download as pdf or txt
Download as pdf or txt
You are on page 1of 18

Ministry of Earth Sciences, Govt.

of India
Indian Institute of Tropical Meteorology, Pune
➢ It is part of the MoES endeavor to provide
world class services to the people of
India.
➢ With 4.0 Peta Flops HPC facility, India is
expected to rise from the 368th position
to around top 30 in the Top500 list of HPC
facilities in the world
Cray XC40 System System Configuration

Cray XC40 18
Cabinets
Blower Cabinets 12
Interconnect Cray Aries with Dragonfly
Network topology
Total Peak 4006 TFLOPS
Performance
Compute Node
Number of Nodes 3315
Processor Intel Xeon Broadwell E5-
2695 v4 CPU
(18 core, 2.1 GHz)
Memory Per Node 128 GB
Total Memory 414 TB
Cray XC40 System System Configuration

Accelerator Node (1%)


Number of Nodes 16
Accelerator Intel KNL 7210, self-hosted
mode
Memory Per Node 96 GB
(GB)
Total Peak 42.56 TFLOPS
Performance
Additional Nodes
External Login 5
Nodes
Utility Servers 8
Utility Racks 1
Cray XC40 System System Configuration

Storage System Configuration


Parallel File System Lustre
Lnet Nodes 39
Storage Array Sonexion 3000
Storage Racks 7 racks + 6 Tape library
frames
PFS – Home
Usable Storage 9.7 PB @ 246 GB/s
Sonexion One MMU + 26 SSU
Configuration
Lustre I/O Nodes 2 MDS + 52 OSS
(Embedded) (Active/Active config)
HDD Data 6 TB – 7.2K RPM on
GridRAID (RAID 6)
Cray XC40 System System Configuration
PFS – Scratch
Usable Storage 986 TB @ 54 GB/s
Sonexion Config One MMU + 4 SSU
Lustre I/O Nodes 2 MDS + 8 OSS (Active/Active
(Embedded) config)
HDD Data 4 TB – 10K RPM on GridRAID
(RAID 6)
Cray XC40 System System Configuration

HSM / Archive Tier

Archive Useable 30 PiB @ 12 GB/s

Storage Array Disk NetApp E5600 series (233 TiB


Cache for HSM @ 20 GB/s)

Tape Library SpectraLogic TFinity

Number of Drives LTO-7 48


Cray XC40 System System Configuration
Software
Cray Linux Environment 6.x
Cray Programming
Environment (CPE) 10 Seats
Intel Parallel Studio XE
Professional Edition 2 Seats
Workload Manager PBS Pro
Additional Debugger –
Allinea DDT & Total View 2048 core
PV-View License 1
Capacity 22KV/433 Volts
Total Capacity 7500 KVA

No of 2500 KVA - 3
Transformer Nos
Purpose For redundancy back

Total
5k KVA
Capacity

No. of DG 1750 KVA - 3 Nos.


Purpose For redundancy back

Total
375O KVA
Capacity

No. of UPS 1250 KVA - 3 Nos.

Battery
15 min.
backup
In the unlikely event
Purpose
of Fire

SYSTEM NOVEC Gas 1230


To cool the Heat
Purpose Generated by the
System

Total
1320TR
Capacity
No. Of
220 TR - 3Nos.
Chillers
Building Monitoring System

To monitor the HPC


Purpose infrastructure health
and functioning
Programming Programming Optimized Scientific
Languages Compilers Tools I/O Libraries
models Libraries

Distributed Environment
Memory LAPACK
setup NetCDF
Fortran (Cray MPT) Cray Compiling
Environment
• MPI (CCE) Modules
• SHMEM ScaLAPACK

Debuggers HDF5
BLAS (libgoto)
C
GNU Allinea (DDT)
Shared
Memory
lgdb Iterative
• OpenMP 4.0 Refinement
• Toolkit
Debugging
C++ 3rd Party Support Tools
Compilers
PGAS & Global
View Intel Composer
• Abnormal
• UPC (CCE) PGI Termination
• CAF (CCE) Processing
Python • Chapel FFTW

STAT
Cray PETSc
(with CASK)
Performance
R Analysis
Cray Trilinos
(with CASK)
•CrayPat
• Cray
Apprentice2

Cray developed Scoping Analysis


Licensed ISV SW
3rd party packaging Reveal
Cray added value to 3rd party
Service Nodes
Compute nodes
elogin Executing the where the jobs
qsub aprun
batch script are executed

Login Servers

Pratyush System includes service nodes, network nodes


and many compute nodes

You might also like