|
|
|
|
|
As technology is moving towards more compute intensive and data heavy applications, demands for a more capable and highly resourceful computers is becoming imperative. There is a continual demand for greater computational speed from a computer system than is currently possible. Areas requiring greater computational speed include numerical modeling and simulation of scientific and engineering problems. Such problems often need huge repetitive calculations on large amount of data to give valid results and computations must be completed within a "reasonable" amount of time. In the manufacturing realm, engineering calculations and simulations must be achieved within minutes or even seconds, if possible. Some areas, such as modeling large DNA structure and global weather forecasting, are "grand challenge problems". A "grand challenge problem" is one that cannot be solved in a reasonable amount of time with today's computing resources.
The continued drive for higher and higher performance system leads us to one simple conclusion .......... THE FUTURE DEMANDS HIGH PERFORMANCE.
Explore High Performance Computing ...
Hardware Architecture
A High performance
computer consists of a number of nodes or independent computing
elements, which work together, interacting with each other to perform
a common task. All the nodes in the cluster are commodity systems - PCs,
workstations, or servers - running commodity software such as Linux.
The sole task of the compute nodes is to execute parallel jobs. In most cases, therefore, the compute nodes do not have keyboard, video and
mouse connected. All access and control to the compute nodes are
provided via remote connections, such as network and/or serial port
through the master node. Since compute nodes do not need to access
machines outside the cluster, nor do machines outside the cluster need
to access the compute nodes directly, compute nodes commonly use private
IP addresses.
HPC Building
Blocks : Software Perspective
A cluster comprises multiple nodes along with its interconnect technology and supporting software.
Computing Platform :
Message Passing based Platforms, each running its own image of the operating system.
Interconnect Technology :
Interconnect technology mainly depends on factors like bandwidth,
latency, reliability, and scalability.
Operating system : Choice
of Operating system depends on its capability to support the existing
software and hardware infrastructures.
Parallel
Infrastructure/Clustering technology : Message Passing Libraries
( for message passing applications), a library of APIs that allow the
coordination, synchronization, and passing of message send data between
the nodes.
* PVM (Parallel Virtual Machine) - Deprecated
* OpenMP for shared memory architecture
* Pthreads for shared memory architecture
* MPI ( Message Passing Interface) for message passing and shared
memory architecture
* MOSIX/OpenMOSIX ( Multicomputer Operating System for Unix) -
combination of message passing architecture and shared memory
architecture.
* data parallel model (f90/hpf)
Cluster management tools : Some of the cluster management tools
are as follows :
* Gangila
* System Imager
* Linux Terminal Server Project (LTSP)
Application Areas of HPC
Some of the major avenues, where high performance computing is being
used, include :
* Bio-Informatics
* Finite Element Analysis
* Computational Fluid Dynamics
* Seismic Analysis
* Genetic Engineering
* Automotive applications
* Aerospace studies
* Chemical and pharmaceutical studies
* Electronics
* Energy research
* Geophysics and oil applications
* Weather prediction
About HPC User Forum
The HPC User Forum is a
new initiative developed by Progression to promote the interests of
high-end HPC (High Performance Computing) users in industry, government
and academia. HPC User Forum is an extension of Progression's new
initiative towards developing a Parallel and Distributed Computing
Community consisting of scientists, academia, researchers, scholars,
industry experts and everyone who is actively involved in these kind of
applications.
HPC User Forum Offers a Forum
Where You Can:
- Discuss ideas and issues with
other high-end HPC users
- Showcase your requirements
and achievements
- Review vendor strategies and
products
- Present consolidated
positions to vendors &policymakers
- Explore new user
opportunities and emerging technologies .
About HPC Team
We believe effective HPC
solutions require good engineering and should be based on customer
needs. An HPC cluster is composed of many independent components.
Selecting the right hardware and software components requires making
informed decisions. Our experience has shown that without proper testing
and benchmarking many solutions are pure guesswork where the likelihood of success is low and the need for additional cost is high. We use our years of experience and benchmark data to determine the best solutions for your needs. We believe that good designs and feasibility studies save you money with hardware, software and deployment costs. We recommend hardware and
software that is right for your applications and needs. When HPC cluster
and applications are designed in this fashion customers can expect the
following benefits:
- Performance - You should get
the best performance at lowest price
- Cost - you should only buy
what you need and nothing more.
- Reliability - What you buy
should work.
We look forward to helping you
achieve your HPC success.
Want to be a part of HPC FORUM ?
Mail us at :
hpcinfo@progression.com
|
|
|
|
