|
Of Knowledge &
Expertise...
Research & Development
is our strength.
We are investigating
a variety of areas, including Operating Systems, High
Performance Computing Systems, Communication Protocols
for Gigabit Networks, High Performance Compilers, Systems
Management Tools, Development Environments and Software
Engineering.
We see ourselves as
a catalyst of innovative ideas in the field of high
performance computing. We innovate through research,
in cooperation with the academia and industry.
The main goal of our
current research is to achieve supercomputer performance
on low-cost clusters for real-life applications with
minimal programmer effort.
The research endeavors,
which are a portfolio of projects aimed at enhancing
CDAC's technical future, can be broadly classified into
the following areas:
Parallel
Programming Environments
Development and execution
of sequential, message passing and data parallel programs
on workstation clusters is a difficult job. Most often
it is difficult or sometimes impossible to successfully
execute long-running applications, as they are typical
for the `Grand-Challenge' problems. So the goal of this
project is to meet this challenge and ease the parallel
application developers job by hiding the complexity
from him and provide him with a flexible and affordable
parallel environment to implement correct and efficient
parallel programs. This will provide the programmer
with tools that help him to develop, execute, debug,
tune and visualize parallel applications.
The current research
issues are:
- Improvement of application
performance through parallelizing compilers
- Performance analysis and improvement
techniques
- Impact on performance by input/output
operations in parallel applications
- Fault tolerance in parallel
programming environment
- Making clusters available
to the scientific community through a meta computing
facility
- Resource management architecture
for clusters Optimization of MPI libraries
High Performance Communication
The advances in network
technologies and cost effectiveness of workstations
has made distributed computing over networks of workstations
very popular. Over the years wide variety of distributed
programming models like MPI, PVM, DSM, CORBA, etc have
evolved and matured. These technology trends have made
it possible to realize a network (or campuswide) metacomputer
comprising of geographically distributed high-performance
machines and workstations connected by gigabit WANs
. So also more tightly coupled workstations connected
by System Area Networks (SANs) called Cluster Computing
Systems can be built. High performance communication
is the key to the success of these systems. Research
at SSDG focuses on exploiting the underlying network
topology and hardware to evolve a new communication
subsystem architectures for both clusters and metacomputers.
Research focus covers all aspects right from algorithmic
optimizations to novel software implementation techniques
like light weight protocols.
High
Performance Scalable Storage
High speed scalable
networks and the communication subsystems have enabled
a closer integration among the components of the cluster.
The research in C-DAC focuses on exploiting this coupling
to evolve new software architectures particularly in
the area of storage subsystems, targeting at greater
performance of the existing interfaces. The research
will also focus on evolving better interfaces based
on the synergy between the software architectures and
the needs of the application classes.
Evolving I/O subsystems
hardware components like NASD, SANs are just an extension
of the storage architecture envisaged in the PARAM cluster,
though it may enable greater flexibility in design.
Thus the research would track the emerging hardware
and software technologies delivering greater performance
to the various application classes.
|
