Slashdot Mirror
World's Largest Working Computing Grid
fenimor writes "UK particle physicists claim that they will demonstrate the world's largest, working computing Grid with over 6,000 computers at 78 sites internationally. The Large Hadron Collider Computing Grid is built to deal with 15 Petabytes of data each year from the Large Hadron Collider (LHC), currently under construction at CERN in Geneva. 'This is a great achievement for particle physics and for e-Science,' says Professor Tony Doyle, leader of GridPP. 'Our next aim is to scale up the computing power available by a factor of ten'."
This week, UK particle physicists will demonstrate the world's largest, working computing Grid. With over 6,000 computers at 78 sites internationally, the Large Hadron Collider Computing Grid (LCG) is the first permanent, worldwide Grid for doing real science. The UK is a major part of LCG, providing more than 1,000 computers in 12 sites. At the 2004 UK e-Science All Hands Meeting in Nottingham, particle physicists representing a collaboration of 20 UK institutions will explain to biologists, chemists and computer scientists how they reached this milestone.
Particle physics experiments at the Large Hadron Collider (LHC), currently under construction at CERN in Geneva will produce around 15 Petabytes of data each year - 15 million, billion bytes. To deal with this vast volume of data, particle physicists worldwide have been building a computing Grid. By 2007, this Grid will have the equivalent of 100,000 of today's fastest computers working together to produce a 'virtual supercomputer', which can be expanded and developed as needed. When the LHC experiments start in 2007, they are expected to reveal new physics processes that were crucial in building the Universe we see today, and shed light on mysteries such as the origin of mass.
Grid computing has been a target for IT developers and scientists for more than five years. It allows scientists to access computer power and data from around the world seamlessly, without needing to know where the computers are. Analysis for particle physics can also be done on conventional supercomputers, but these are expensive and in high demand. Grid computing, in contrast, is constructed from thousands of cheap units that can be increased to meet users' needs. Like the web before it, the Grid has the potential to impact on everyone's computing.
GridPP, the UK's particle physics Grid project, was set up by the Particle Physics and Astronomy Research Council in 2000. On 1 September this year the project reaches its halfway point, with the official end of its first phase and the start of GridPP2. According to Dr Dave Britton, the GridPP project manager, "The first half of the project aimed to create a prototype Grid - which we've done very successfully. Having proved that a Grid can work, we're now focussed on developing a large-scale stable, easy-to-use Grid integrated with other international projects. This will let scientists tackle problems that are much larger than those possible today."
Dr Jeremy Coles of Rutherford Appleton Laboratory is the GridPP production manager, responsible for making sure the Grid works on a day-to-day basis. He is giving the main GridPP talk in Nottingham, and stresses, "There are a lot of challenges in front of us as we expand our production Grid. In addition to the technical problems involved in providing a well-monitored, stable Grid, we need to address wider issues, in particular encouraging an open sharing of resources between groups of users."
In Nottingham, conference delegates will be able to see how the particle physics Grid works. GridPP has developed a map that shows computing jobs moving around LCG in real time, as they are distributed to the most suitable sites on the Grid, run their programmes and then return their results home. The map can be seen here. Dr Dave Colling, from Imperial College, London, whose team built the map, said, "It can be difficult for people who have never seen a Grid working to imagine what it does. Our map is an easy way to see how a Grid can let scientists use resources all over the world, from their desktop. It's also useful for experts, who can easily see how well the Grid's working."
Professor Tony Doyle, leader of GridPP, explained, "This is a great achievement for particle physics and for e-Science. We now have a true international working Grid, running more than 5,000 computing jobs at a time. Our next aim is to scale up the computing power available by a factor of ten, so that we'll have 10,000 computers in the UK alone, ready for the Large Hadron Collider in 2007"
In Grid computing, which is not exactly the same as high-performance computing, the number of flops doesn't really matter that much, it's more about providing an environment for multiple users to address problems that can be solved by splitting it up in a huge number of smaller tasks.
your joke being funny not withstanding, that's a map of America, probably the least relevant place to show for this particular project.
CERN and Grid is European, notably Switzerland, France and UK.
the USA has plenty of great particle physics of its own (excitable New Yorkers beware - there's a particle accelerator on your doorstep - think of the children!) but this is not one of them.
I think the joke works better as a reference to the Hitchhiker's Guide to the Galaxy.
That writeup looks a lot like the one at The Register -- which came out a good two days early, the same day the results were actually announced at the AHM conference.
The LCG resources have several different things that most home machines do not:
1) A Linux install with the requisite libraries for the already-written experiment analysis programs to run on.
2) Fast network interconnects, both to other LCG cluster nodes at the same site (using Myrinet, Infiniband, etc.) and large network connections to other participating sites (ie 100Mbit+).
3) Large amounts of reliable local storage, ie 1TB+.
SETI@Home-like distributed computing problems only work well for problems which do not require large amounts of communication between nodes before, during, and after an individual run. Many problems do not fall into this category.
[80s]So funny I forgot to laugh[/80s]
People are mostly using "Scientific Linux", an in-CERN-house Redhat fork, but some sites are experimenting with other stuff - one of the computer science aspects of the grid is researching how to make good use of heterogenous systems, though different linux distros aren't amazingly heterogenous in the grand scheme of things, there are challenges.
And yes, there are people working on Gentoo, believe it or not.
And Debian and fedora core 2 and 3 and mandrake clic and suse.
And IRIX, AIX, Solaris and Mac OS X from the Real Unix camp.
And there's some poor fuckers somewhere working on windows too.
I wonder how they are coordinating the use of all of those computers? The article doesn't say that they will be exclusively for this project
Well, GridPP is exclusively particle physics, although there are other grids in construction. Large numbers of people will do large numbers of analyses with the LHC data - it's not just a case of running one job on all the data, it's a case of many jobs and many subsets of the data.
Globus and digital certificates are also part of your answer.
So right now it isnt even 100 thousand computers, maybe not even close, so the computing power might be similar. (assuming 15 thousand active computers on d.net)
The point is not so much assembling all that computing power now (the LHC won't come online till 2007 or so anyway, and you don't really need the Grid to run Monte Carlo) so much as assembling the infrastructure so that when 100 universities go out and buy new analysis farms in 2007, they can get tied together and used efficiently.
lhc at home is not for processing the data output, but helping them to position the magnets as they
*build* the LHC.
Because the Earth is a LOT brighter than the stars (because the stars are far away), and to properly expose the Earth onto whatever media is being used(film, CCD, whatever), less exposure is needed than would be necissary to pick up any stars (save the sun).
Just like pictures from the moon - you'll not see any stars in pictures taken of the moon on the moon (by Neil Armstron et al).
Hope that helps
I was under the impression that the world's largest working grid was the United Devices grid.org project. They have over 2.5 million registered users and average over 300,000 work units returned every day. check out www.grid.org They are also doing real science.