What is Grid Computing? How can it be important to mechanicians? When?

A few years back, grid computing was hot.  There were many conferences dedicated to the technology and applications, and there are even many grids, actually the new names of the known supercomputing centers.  Since then, it has been quite and I did not pay much attentin in last two or three years.

The concept of grid computing is a natural extension of the ubiquitous internet connection and the increasing bandwidth.  If the connection can be utilized and the resources can be assembled, it is just a shared service in the Internet era.  Unfortunately, just like the tiem sharing service (TSO) never took off, the grid computing has also fall into the same trap.  For those who need computing power and resources, they always have it.  Look at the new projects and computer clusters in the national labs.  For some of user who demand more, they cannot afford the grid service either.  Make the grid work needs a lot of investment for the strealining, coordinating, and retooling.  This may be the reason why the NSF pulled the plug for more money on the grid computing.

The existing grid are either empty or idling.  The grid is a good concept, but make it work needs more.

Most computing intensive research can be done with the existing supercomputing centers and government labs.

Grid will have a long time before it can be reached by all users in a seamless manner.

Consequently, the grid might have very little effect on mechanicians.

At some point, I was wrong.

I thought grid computing was simply another name for parallel computing

on cluster. Then I visited in China most likely a state key laboratory where they explained to us the meaning of grid computing.

I remember it as a way to connect distributed people.

As an example, the governement was using it as a way to

interact on SRAS with researchers in China. The topic was too delicate

to discuss it for instance on radio.

I like this idea of connecting distributed people.

Isn'it what imechanica is alla bout : distributed knowledge to distributed people.

Merry Christmas to you all.

Nicolas

Nicolas MOËS

Ecole Centrale de Nantes

1 rue de la Noë, 44321 NANTES

FRANCE

An interesting article from Chicago Tribune
"Grid computing gaining acceptance"
by Jon Van Published January 2, 2007

The GRID is an aggregation of computational resources where computing is performed over a network of pooled assets. One can think of these assets as a virtual pan global organization where resources are heterogeneous (platforms, software, servers, mainframes) and autonomous.

Perhaps the most famous examples of scientific computing on a GRID is the SETI @ HOME program (Search for Extraterrestrial Life), which connects 1.2 million computers and has the ability to compute over 200 Teraflops, and the Large Hadron Collider Computing GRID which has 100 sites in 31 countries which contribute 10 000 CPUs. Singapore is home to one of its clusters. The TeraGRID (http://www.teragrid.org/about/) in the US is an open scientific infrastructure with nine partner sites. Using high speed network connections, researchers can access over 100 Teraflops of compute power and petascale levels of storage capacity. National GRID efforts also exist in countries like UK, Singapore and Cyprus.

So GRID computing has become a reality and its philosophy (federating resources) is being embraced even by those outside academia and research. Companies like SUN, HP, IBM, Oracle (and ABAQUS) now recognize a new business model of commerce. Marketing services over distributed resources. Which means software and services can be transacted on a utility basis. Pay as you use. Attractive to enterprises because the total cost of ownership reduces when peak demands can be purchased automatically. Seductive to the scientific community because cost of licenses and compute time is less, and diversity of software exist on demand.

For the scientific community, the GRID is much like the internet in the early prospecting days. An opportunity to devise new ways of doing things. Like the internet’s evolution, the scientific community is blazing the trail of developments and use with commerce fast following suit. The platform which rides on the GRID is often open sourced Linux based. Therefore receptive to all types of codes. For mechanicians it is perhaps, a canvas to work on problems previously limited by resources. Problems with higher degrees of complexity, multiscale and multiphysics modelling and, computations on highly coupled systems. Germane for systems in biology where the symphony of organization and differentiation is integral to understanding associated phenomenon. Another enticing opportunity is the ease for groups at various geographical locations to contribute to a certain problem over a GRID. Contributing to needed multidisciplinary expertise otherwise limited at one place or team. GRID computing thus offers incredible compute resources and community-based interaction at one’s finger tips. We use what we only need and unbridle imagination.

Raj. Thampuran, Executive Director, Institute of High Performance Computing (IHPC), Agency for Science, Technology and Research (A*STAR), Singapore.

When the Large Hadron Collider at CERN comes into  service sometimes next year, it would produce 12 penta bytes (12PB) of data per year. Forschungszentrum Karlsruhe   is one of the partners for analysing this massive amounts of data. In this regard the German Government is acively encouraging the setting up of D-Grid.

Just to give an idea:

1PB data can be transferred on an ISDN line in 6000 years and on a DSL line in 300 years. If the data transfer speed is around 10GBPS then 1PB of data can be transferred in 10 days!

V. Hegadekatte, University of Karlsruhe, Germany