The hyper-dense blade designs already on the market from IBM rivals lack the power to perform in an application or database environment, Tim Dougherty, director of server blades at IBM, said.
The company's server blade hardware and software architecture, dubbed IBM eServer BladeCenter, will be made up of high-performance, two-way server modules running Intel's Xeon MP processors as well as IBM's own Power chips, Dougherty said.
"Our vision is a vision of server blades that don't force customers to sacrifice performance and reliability in order to jam as many servers into a rack as they can," Dougherty said. "We want to offer a high-performance blades system and prevent this kind of ill-advised trade-off between footprint and performance."
Early blade offerings from startups such as RLX and Racemi were predominately targeted at simple, appliance-like, front-end Web server tasks.
More recent server blade offerings from Hewlett-Packard and Compaq appeared with roadmaps that add processor power to the blades and evolve them into the application and datacentre layers.
But IBM plans to start at the high-performance level with BladeCenter, Dougherty said. The first BladeCenter products will be dual-processor servers which will fit 84-to-a-rack and incorporate gigabit Ethernet as an I/O backplane for each of the server blades, he said.
Making it possible to build a complete network from IBM's BladeCenter, the company will offer a combination of server blades, IBM TotalStorage blades, and a variety of networking blades.
IBM Director software will provide the management tools for BladeCenter and will include elements of IBM's task automation and self-healing technologies such as eLiza and Project Oceano, according to IBM.
IBM will also spearhead a BladeCenter Alliance Program that will foster interoperability between BladeCenter and products from Broadcom, D-Link, Intel, Microsoft, Nortel, QLogic, and IBM's own Lotus and WebSphere products.