| Introduction |
|
 C-DAC
is India's national initiative for building capability
in the area of High Performance Computing systems and
solutions. Towards fulfillment of this goal, C-DAC has
advented the OpenFrame Architecture which has been realized
in its celebrated PARAM series of High Performance Computers.
These systems are scalable from desktop to teraflop
range embodying the emergent industry-standard building
blocks. PARAM Padma is the latest in the PARAM series
and is a result of its third mission project in High
Performance Computing.
To provide high
performance computing facilities and services for the
scientific & research community and for the enterprise,
C-DAC had earlier set up the National PARAM Supercomputing
Facility (NPSF) at Pune, housing its PARAM 10000 supercomputer
with a peak computing power of 10 Gflops. In its continuing
effort in this direction, C-DAC has now established
C-DAC's Terascale Supercomputing Facility (CTSF) at
C-DAC Knowledge Park, Bangalore with the PARAM Padma
machine in its Terascale configuration. The PARAM Padma
at CTSF has a peak computing power of 1 Teraflop and
5 TB of storage and it has achieved 532 Gflops on the
LINPACK benchmark placing the site at number 171 in
the June 2003 list of Top 500 fastest computers in the
world. This is the first time an Indian organization
has found a place in the list displaying our technological
competency in the arena of supercomputing.
Facilities such
as CTSF have to run with highest uptime and adequate
technical support to provide quality service to its
users. This requires appropriate power back-up with
UPS and DG sets, and spares for technical resources
all the time on a 24/7 basis. Also, system administration
and technical consultancy support are an integral part
of running such a facility. Further, due to fast obsolescence,
constant upgradation of the technical (both hardware
and software) resources is needed to stay current. In
order to accomplish these, C-DAC has deployed considerable
resources to offer a quality facility on a 24-hours
uninterrupted basis. These have, therefore, to be at
least partly recovered through its usage to sustain
the operation. Keeping in view the practice followed
worldwide, the Technical Affiliation Scheme has been
introduced with the objective of encouraging the potential
users of high performance computing resources in a cost-effective
way. The scheme will also provide a state-of-the art
technically congenial environment to the researchers.
This scheme provides
supercomputing facility to various industries as well
academic and R&D institutions to cater to their
diverse applications, which require supercomputing power.
For a long time it has not been economically viable
for the industry and academic/R&D institutions to
use High Performance supercomputers. This scheme is
a step in the direction to address this problem by providing
state-of-the-art "PARAM Padma" parallel machine
to external users. For this, an interested user has
to be enrolled as a Technical Affiliate of CTSF in order
to use the computing resources of this facility. They
can then use CTSF resources on the terms and conditions,
details of which are given in this booklet.
|
| 1.
The Technical Affiliation Scheme |
|
Facilities such as
CTSF have to run with highest uptime and adequate technical
support to provide quality service to its users. This
requires appropriate power back-up with UPS and DG sets,
and spares for technical resources all the time on a
24/7 basis. Also, system administration and technical
consultancy support are integral part of running such
a facility. Further, due to fast obsolescence, constant
upgradation of the technical (both hardware and software)
resources is needed to stay current. In order to accomplish
these, C-DAC has deployed considerable resources to
offer a quality facility on a 24-hours uninterrupted
basis. These have, therefore, to be at least partly
recovered through its usage to sustain the operation.
Keeping in view the practice followed worldwide, the
Technical Affiliation Scheme has been introduced with
the objective of encouraging the potential users of
high performance computing resources in a cost-effective
way. The scheme will also provide a state-of-the art
technically congenial environment to the researchers.
This scheme provides
supercomputing facility to various industries as well
academic and R&D institutions to cater to their
diverse applications, which require supercomputing power.
For a long time it has not been economically viable
for the industry and academic/R&D institutions to
use High Performance supercomputers. This scheme is
a step in the direction to address this problem by providing
state-of-the-art “PARAM Padma” parallel machine to external
users. For this, interested user has to be enrolled
as a Technical Affiliate of CTSF in order to use the
computing resources of this facility. They can then
use CTSF resources on the terms and conditions, details
of which are given in this booklet.
|
| 2.
Resources at CTSF |
|
Following are the resources currently
available at CTSF. These are continuously upgraded to
address the requirements as they emerge.
|
| 2.1
Hardware Resources |
- Param Padma Compute Nodes :
| Configuration: |
54 Nos. of 4 Way
SMP nodes and 1 No. of 32 Way SMP node |
| No. of Processors:
|
216 (Power4 @1GHz)
& 32 (Power 4 @1.1GHz) |
| Aggregate Memory: |
0.5 TeraBytes
(@ 8GB per node and 64 GB for large SMP node) |
| Internal Storage: |
4.5 TeraBytes
(@ 72GB per node and 576 GB for large SMP node) |
| Operating System: |
AIX 5L / Linux |
| Aggregate Peak
Computing Power: |
1005 GFs (~1 TF)
|
- File Servers:
| Configuration: |
6 Nos. of 4 Way
SMP |
| No. of Processors:
|
24 (UltraSparc-III@900MHz) |
| Aggregate Memory: |
96 GigaBytes |
| Internal Storage: |
0.4 TeraBytes |
| File System: |
QFS |
| Operating System: |
Solaris 8 |
- Networks
| Primary: |
PARAMNet-II @
2.5 Gbps Full Duplex |
| Backup: |
Gigabit Ethernet
@ 1 Gbps Full Duplex |
| Management: |
10/100 Mbps Fast
Ethernet |
- External Storage
| Storage Array: |
TeraBytes with
16 T3 disk arrays |
| Tape Library: |
12 TeraBytes -
(5 LTO drives) |
| Storage Area Network
consisting of 2Gbps FC 16 port switches and
a 1 Gbps brocade 2800 FC switch connected through
a fibre channel. |
- " Two Compaq Evo W6 graphic workstations running
Slackware Linux 9.0 and Windows 2000 and number of
user terminals on the LAN to connect to the PARAM
Padma Cluster.
- Laser printer.
|
| 2.2
Software Resources |
| 2.2.1
Operating System |
|
The Operating System of PARAM Padma
is AIX 5L which supports multithreading and multiprocessing.
This System V-style Unix provides a familiar development
environment to the users. Optionally, Linux operating
environment is also provided.
|
| 2.2.2
Compilers and Related Tools |
|
The following compilers and related
tools are available on PARAM Padma.
- XL Fortran compiler and Run Time Environment
- F90 Compiler
- F77 Compiler
- C for AIX V 6.0
- Visualage C++ Compiler
- CDF90 Compiler - C-DAC's Parallel Fortran Compiler
- C-F77to90 - C-DAC's converter for Fortran77 to Fortran90
programs
- Complete gcc & gnu libtools
|
| 2.2.3
Parallel Processing Libraries |
- Message Passing Interface (MPI) - IBM PE (Parallel
Environment)
- Public domain MPI - MPICH 1.2.4 for XL & gcc,
g77
- C-MPI - C-DAC's optimized implementation of MPI
for PARAM Padma
- KSHIPRA - C-DAC's Communication substrate for PARAMNet
II
|
| 2.2.4
Parallel Numerical Libraries |
- IBM ESSL & PESSL
- SCALAPACK
- LINPACK
- PLAPACK
- MASS
- PetSC
|
| 2.2.5
Program Development Tools |
- DIViA - C-DAC's parallel program debugging environment
- Metric Advisor - C-DAC's tools for enabling developers
to build run quickly.
- dbx, xprofiler
|
| 2.2.6
System Administration and Job Management Tools |
- Cluster Monitoring Tools
- System Accounting Tool
- Dedicated Slot Booking Tool
- Load leveler 3.1
- PARMON - C-DAC's comprehensive cluster monitoring
system
- Veritas Backup Tool (Net Backup 4.5)
|
| 2.2.7
Visualization Tools |
- Khoros
- Myavi
- OpenGL
- CWP
- Rasmol
|
