Big Blue to show off upcoming server architecture

IBM will demonstrate a new architecture for Intel-based servers Monday, all part of Big Blue's plan to bring more complex computing techniques to the corporate masses.

At its PartnerWorld convention in Atlanta, IBM will show off the first servers containing "Summit," the code name for a chipset that will come out commercially toward the middle of the year.

Summit will allow IBM to incorporate more of the features typically associated with its high-end server lines into its generally less expensive Intel-based servers. By migrating these features to its Intel line, Big Blue can differentiate itself from competitors that perform far less research and development.

The chipset, for instance, will permit corporations to put up to 16 processors in a single machine at some point this year, while many IBM competitors will be stranded at four processors, said Tom Bradicich, director of architecture and technology for IBM's xSeries line of Intel servers.

The server consists of as many as four processor "quads." Each quad has four processors, its own memory and input-output abilities. The quads can communicate over a high-speed "scalability port," a technology that came from IBM's 1999 acquisition of Sequent.

"Summit architecture puts us in a unique position," Bradicich said.

The processors will also be capable of being partitioned. That is, some processors can be dedicated to running a Windows database while others perform tasks on Linux. Customers will also be able to "hot swap," or add and remove, memory while the server is running.

The chipset will first be featured in servers containing "Foster," a version of the Pentium 4 for servers coming toward the middle of the year. The same basic chipset will then be used in servers containing "McKinley," a 64-bit chip coming from Intel in 2002.

"One component of the chipset changes to accommodate McKinley," Bradicich said.