It's supercomputing - It's parallel computing - It's meta-computing - No, it is The GridAmsterdam 05 March 2001 "Where the ??**!?! is my program? I thought I did send it to the US to do some calculations for me, but it seems it got to Czech Republic on its own and is now stuck somewhere in Prague!". Science fiction? No it was scientist in the chair next to me who was attending a session at the Global Grid Forum meeting in Amsterdam. Well, he was following two lectures at the same time, and monitored the activities of a program he did send off from the computer at his department in Poznan Poland. His laptop was connected via a high-speed wireless connection to the Internet. Over the Internet he followed a live video broad cast of the lecture in another room of the conference. The program he did send off from his machine in Poland from another window on his laptop, was living on the Grid, a new Internet based environment, where programmes can move themselves to computers that have time available to run it. Whether that computer is in Europe or Australia, the programme does not care, it just goes to where the best resources are available. When it is kicked away, it just find another place on the Grid.
What is the Grid? You may heard of it, but no, it is not the next generation Internet. It is not providing you access unlimited computing power,
and unlimited resources, any place anytime. At least not for a very long period to come. What it is is a combination of very fast
networks and a number of tools that will help scientists to share data and resources and collaborate over their Internet. The example of the Poznan scientist
is a preview of what will come in the next years.
Researchers over the world have access to networks much faster than most of us. Most national research networks run at Gbit/s speed already in will be
upgraded to 80 or even 100 Gbit/s per second in the coming years. Compare this with the "broad band" access of 1 Mbit/s that is available to (part of the) European
citizens through cable or xDSL connections. Having so much bandwidth makes people think differently. Doing number crunching for instance. Traditionally you
take a large computer with enough memory and disk space. You move the data and the program to it and than you execute it on that machine. With enough band width
you can get the data in during the calculation. Or you can do part of the calculation on one machine, and the remainder on another. This is not "new" technology.
There have been experiments going on for many years. But then it was called meta-computing, because the Grid was not invented yet. What is new is the
scale of usage of these techniques made possible by the much faster networks.
It is not only meta-computing that finds it way to the Grid. Everyone active in the high-performance computing arena, be it in cluster computing, large databases,
collaborative work, or virtal laboratory, or even monolitic parallel computing, is now declaring that work as Grid computing avant la lettre.
The Internet itself made it easy to access computers anywhere, anytime. Of course you have to have rights to access the machine, but then, there is not much difference
with using a local machine. Of course the network connection has to be fast enough. For researchers this opens the possibility to use programmes on other machines,
because the other computers are more suited for a certain problem, or because there is special software available. But before the Grid, you had to have accounts
on all machines, and take care that when you did send a program over, you had the right account. Grid technology automates this: you can enter your registrations
and accounts in a directory, or they can be assigned automatically, for instance because you participate in a project. Then it is possible to send of a program that wanders
around on the Grid from computer to computer looking for the right resource it needs. Whether the techniques that will be developed for the Grid will usable outside
the scientific arena remains to be seen.
Another development in the scientific world is the proliferation of very large date generation experiments. Particle colliders for instance, or satellite data,
they are all generating streams of terabytes of data. These amounts are so huge that even with the fast networks, they can not easily be moved. Hence they have
to be analysed in an intelligent way close to the source and only smaller data sets send to research groups. On the other hand, there sometimes are hundreds of researchers
working on one experiment. Giving them fast and seamless access to data from anywhere in the world, makes it possible for them to stay at home and work
on the experiments from their own institute. This is cheaper than moving them all to the experiment. In some cases, it would even be impossible. For instance with
instruments high in the mountains or in polar areas. Within the Grid this seamless access will be provided by applying a number of software-tools and data management
tools experimented with during the past years.
As usual in the research world, it is not always the best software tools that are used. The Grid is no exception. One of the reasons is that researchers do not like to pay for software.
They are producing loads of specialised software themselves. So why pay for basic tools? On the other hand it is often difficult to change or add features to
commercial software. Public domain, open source tools, although not having the same robustness and quality as commercial tools are often preferred. Although
this can be a viable strategy in research, it makes a one to one copy of academic Grid results to the main stream Internet world almost impossible.
But undoubtedly results will spring off the Grid that will be beneficial in a broader area. Imagine replacing the program from the Poznan scientist doing serious scientific work with, for instance, a program representing a game figure. The game is to chase other game figures over the world moving from computer to computer. Of course you are not in the lecture room but are attending a soccer game and the person you watch streaming full screen to a window on your laptop is your wife, or is not she?
Ad Emmen |
 |