IBM launches supercomputing on demand
Linux cluster, AIX grid
IBM Corp's Server Group and Global Services organizations have got together to launch the first true utility computing offering for server customers from Big Blue, writes Timothy Prickett Morgan. The supercomputing on demand offering is the first in what will likely be a flood of similar announcements for just about every kind of workload and customer.
Under the program, IBM will build supercomputer application hosting centers around the world based on its AIX-based pSeries line and Linux-based xSeries Intel servers. The first supercomputing on demand hosting center will be located in Poughkeepsie, New York, and it will comprise a grid of hundreds of eight-way pSeries 655 midrange machines running AIX and what IBM calls a massive Linux cluster based on its xSeries 335 1U and xSeries 345 2U two-way, rack-mounted servers, which are powered by Intel's Pentium 4 Xeon processors. IBM will eventually set up other supercomputing on demand hosting centers around the U.S. and in other geographies, and these will all be linked using the open source Globus grid software.
The exact configurations of this grid are not available since the first customer for the on demand service - the PGS Data Processing division of Petroleum Geo-Services, an oil exploration and visualization firm that is based in Lysaker, Norway and Houston, Texas - does not want to let its competition know what iron it is using.
PGS is already running its own Linux clusters and will move those Linux applications to the IBM service, and says that it will save $1.5m a year in costs since it doesn't have to buy, configure, maintain, or upgrade that Linux cluster, and more importantly, it will only pay for the data processing capacity it uses. Companies like PGS often have peak computing demands that are far in excess of their average demands, and they have to acquire equipment to pay for the peaks or risk losing business.
In this regard, all IT organizations face similar problems. This is why IBM is so excited about its so-called e-Business On Demand initiative, spearheaded by chairman and CEO Sam Palmisano as the defining characteristic of his tenure as the head of the company.
While the utility computing model is something that all IT vendors are trying to move their customers to because it takes some of the risk out of predicting sales and tends to lock customers into tighter deals, the question that no one has been able to answer is this: Exactly how, in principle, does On Demand computing differ from the application service provider model that was supposed to take over the corporate computing world a few years ago? The technology might be evolving with grid and Web services middleware, but most companies are still buying servers and software.
Should this on demand initiative take off, IBM will probably have to consider installing supercomputing equipment from rival vendors. IBM Global Services has long since supported competitive equipment, and while it doesn't brag about it, the organization is fairly agnostic about server platforms so long as customers are willing to pay for services and support on the gear.
Exactly what Global Services is charging for the supercomputing on demand offering is unclear. IGS does not publish a price list, in contrast with the Server Group, and IBM sources say the pricing will depend on the type of iron used, how much iron customers need, how much processing capacity they use, the workload they have, and how much work IBM needs to do to support that application as it runs.
On demand computing has not evolved to the point that pricing and competitive comparisons can be made as is possible with Web and email hosting, for instance. And because of the complexities of corporate application workloads, that may never happen.