Tomorrow’s service oriented infrastructure (SOI – see pages 16 – 17) will be composed of an infinitely versatile pool of shared resources that can be rapidly provisioned and reconfigured to meet changing demand for different services. Much of this flexible new infrastructure will be built out of blade computers.
Blade computers first appeared as a response to the dot-com era’s urgent need to cram more servers into their rapidly filling data centres. Pioneering vendors like Egenera and RLX Technology solved the problem by going back to basics.
Servers were whittled back to the bare essentials and reinvented as slender “blades” which treated once dedicated resources as shared components accessed via a high-speed bus, housed in a dedicated enclosure.
This approach not only allowed data centre operators to increase processor density to unprecedented levels, it also handed them a simplified, and more flexible unit of processing power: a design that helped reduce the complexity in the data centre, and made it easier for servers to be installed, expanded or replaced.
Blades have since become part of the fabric of many major data centres, and blade architectures have found their way into the product portfolios of all the major server vendors, including IBM, Hewlett-Packard, Sun and Fujitsu-Siemens. They are still far from being close to entirely supplanting conventional servers, and last year, according to Gartner, still only accounted for 7% of total server sales.
However, it is clear that blades have already usurped traditional servers as the chief focus for most of the advanced and innovative engineering breakthroughs in the server space.
Earlier this year, for instance, Hewlett-Packard launched its c-7000 blade server enclosure with the promise that it would “reinvent the data centre” – reducing the cost of new server acquisition by 41%, cutting operational costs by 60%, and reducing the average time taken to provision a new server by 96%.
HP was able to make these claims because the c-7000 is an example of what Gartner calls the “fourth generation” blade server: an altogether more sophisticated product than its space-saving predecessors.
As HP’s claims make plain, the modern blade system has less to do with allowing more power to be squeezed into less space (although that is still a key advantage of blades) and a great deal more to do with answering the need for more fine-grained control of their physical resources.
This will include providing IT infrastructure builders to more easily mix and match different categories of processing component. Although early blade servers were truely bare-bones devices, accommodating just a single commodity processor and little else, today’s blades offer a greater choice.
Customers can still choose to simply cram dozens of simple mono-processor blades into a single chassis to provide a scalable platform for a customer-facing web server.
However, if their needs are more complex, they now have the choice of populating the chassis with dual or quad processor blades; blades with large maths co-processor or floating point engines; or blades with onboard memory and storage.
In essence, customers can choose to build their own computer systems out of blades and, by so doing, do a better job of tailoring back-end resources to the precise needs of their business users.