09 Jul 2008
IBM’s latest supercomputer, said to be the first to deliver a petaflop of performance a million gigaflops is a mine of mind-boggling statistics. Named Roadrunner after the official state bird of New Mexico, the system uses over 6,000 AMD Opteron chips combined with over 12,000 IBM Cell processors.
The system has 98TB of memory, and is separated into 17 “connect units”, which are blocks of 180 compute nodes, with each node consisting of two dual-socket Cell blades and a single Opteron blade.
Roadrunner cost £60m to build, covers 6,000 square feet and weighs over 220 tons. Later this summer, it will be loaded onto 21 trailer trucks and transported to the National Nuclear Security Administration at the Los Alamos National Laboratory in New Mexico. Here it will be used for nuclear weapons safety and reliability simulations, to make sure that US nukes only go off when they are supposed to.
If you are curious about what operating system Roadrunner uses, the answer is Red Hat Linux. In fact, most high-performance computing (HPC) systems use non-Microsoft environments, a state of affairs that the software giant has been trying to change for years.
At the recent International Supercomputer Conference in Dresden, Microsoft announced a new release candidate for its HPC Server 2008 package, and revealed that it had managed to break into the list of the world’s top 25 supercomputers, coming in at number 23 with a system used by the US National Center for Supercomputing Applications.
So why is Microsoft lagging behind? Well, it may be because firms normally use HPC systems in highly critical parts of their operations. For example, energy companies use them in prospecting for new gas fields, while pharmaceutical firms use them in designing new drugs.
In scenarios such as these, every second of compute time is valuable, so when system problems occur, the faster the fix the better. The idea of having to get on the phone to Microsoft tech support does not sit too well with those whose business it is to keep the numbers crunching. But with Linux systems, anyone can have access to the source code, which means problems take less time to resolve.
But demand for Microsoft HPC kit is finally beginning to pick up. Viglen chief executive Bordan Tkachuk said that his company is definitely seeing more interest in the technology.
It seems that firms that already use Windows on their corporate networks are now more easily swayed into using Microsoft’s HPC platform. Demand from financial firms is said to be particularly strong. Apparently, many of the institutions most responsible for the current credit crunch are now installing HPC systems to perform risk analysis in an effort not to get caught with their trousers down next time.
The builders of HPC systems are thus among the few to benefit from the problems currently affecting global financial markets. Some financial firms are even analysing historical data in an effort to pinpoint why they got caught out.
You have to wonder, though, whether it really requires an expensive piece of HPC equipment to work out that selling over-priced houses to people who can’t really afford them and dishing out credit cards like confetti maybe isn’t the most sustainable business model.
This is a very narrow view of the HPC world. Time to decisions in business (not just research) is connected to way more than just the raw compute time.
For example the preparation of the whole computation and the collaboration and discussion with other people before that is fundamental.
The ability to submit job easily and seamlessly directly from inside the desktop apps where most of the innovatin and decision making is done is also fundamental.
Having IT resources able to setup and manage the IT is also very important.
The new workstations will have a ton of compute capacity that will be used to make faster decisions every day and I submit that the combined business effect of that is much bigger than the effect of one single monster computer.
So the important distinction is between high end science HPC and business and daily engineering HPC. Both are needed, but Linux did not and cannot seamlessly extend and integrate with the core of the business tools that is 90% based on Microsoft technology. Just remember that the vast majority of business in the world runs on Excel and that is where th vast majority of daily computation and related decisions are done.
A lot of business and people in the world do NOT use HPC every day exactly because it has been "different", complex, difficult to use and manage.
Microsoft will finally help bring HPC to everybody in a cost effective and easy to use way, not just with Windows HPC but with the whole collaboration, knowledge management, OBA and mashed apps entering the engineering and science world.
Posted by: Michele 11 Jul 2008
Have your say on this article
Newsletters
Latest stories from Applications
You may also like
Applications jobs
Technology Patent Wars
Case studies from large organisations across all sectors
... And rich media, and flexible working, and peaks in traffic ...
Upcoming Events
Join us for this Computing web seminar, in which the Head of BI at the Co-operative Group Nick Colebourn will be explaining just how he reigned in the Group’s sprawling database estate and how significant savings were realised and data quality improved as a result.
Date: 31 May 2012
Time: 11:00 AM
Live June 13th 11:00am: Register now. During this web seminar we will be looking at the sorts of incidents that can bring data centres grinding to a halt and what can be done about them.
Date: 13 Jun 2012
Time: 11:00 am
Receive the latest jobs direct to your inbox
Are you being paid what you are worth?