The winds of change may turn bitter now that three of the largest server companies have announced plans to go over Intel Corp.’s head to develop their own I/O standard.
The formation of the Future I/O forum, consisting of IBM Corp., Compaq Computer Corp. and Hewlett-Packard Co., as well as some peripheral companies like Adaptec Inc. and 3Com Corp., was announced just days after Intel’s Next Generation Input/Output (NGIO) group was made public.
The NGIO group of companies, including Dell Computer Corp., Hitachi Ltd., NEC Corp., Siemens Information Communication Network Inc. and Sun Microsystems Inc., was formed in order to replace the PCI and the PCI X specs in enterprise servers. The industry widely views a new architecture as necessary because the existing standards are not keeping up with faster processor speeds or compute-intensive areas like electronic-commerce.
Jonathan Eunice, analyst and IT advisor at Nashua, N.H.-based Illuminata Inc., said having two specifications co-existing cannot be good for the computer industry.
“It could be a major battle — and one that could be very dividing and nasty for a lot of people,” he said. “That kind of thing can create a major schism, and it can be quite painful for a large part of the industry.”
Doug Cooper, Canadian marketing manager for Intel of Canada Ltd. in Toronto, said he hopes all the key industry players can work together to form one I/O standard.
“Our view is that no one company can drive a specification through to become a standard — including Intel; we really need the support of the broader industry,” he said.
Future I/O members say the reason they are going ahead with their own specification is because Intel’s offering is not robust enough, but Eunice feels it is a business issue more than a technology problem, and that many companies are tired of so much of the profit and control going to companies like Intel and Microsoft Corp. NGIO’s stipulation that no royalties will be paid to members for technical contributions is one cause for the rift, he said.
According to Eunice, Intel will have trouble enforcing NGIO without the support of HP, Compaq and IBM. On the other hand, going along with Intel’s proposal could have its advantages, he said.
“It is very hard to set an industry-wide standard. Even though many people curse Intel or curse Microsoft for setting all these standards, nine times out of 10 having an organization that is powerful enough to set a standard is beneficial to everyone, because then everyone can get on the same boat.”
Ken Jansen, director of advanced architecture and design for Compaq Computer Corp. in Houston, said Future I/O was developed to protect customers from suffering through multiple revisions of a specification once it is in place.
“We are not in a mad rush to put something out before it is ready. Customers tell us they don’t want to go through the ‘spec of the month club.’ They want it to work reliably and they want it to work for a long time,” Jansen said.
But, according to Eunice, there are few differences between the two proposals. “All the concepts of serial interface — point-to-point links with switched-fabric routers, with various types of fault tolerance, fault resilience and multi-pathing built in – all those technical concepts are identical.”
Dr. Tom Bradicich, director of architecture for Netfinity Servers at IBM Corp. in Research Triangle Park, N.C., admitted the two proposed specifications are actually quite alike.
“They both use the IBM 390 I/O architecture as a reference, but we have an advantage because we know how to put those together from a systems level so it’s vastly less costly and less complex.”
NGIO is scheduled to be released sometime in 2000, but Future I/O will not arrive until about 2002. Bradicich accused Intel of rushing the specification development process. “But we are not going to do that with Future I/O,” he said. “It is much too important.”
Martin Whittaker, research and development manager for Enterprise NetServers at Hewlett-Packard Co., in Cupertino, Calif., also has a problem with Intel’s targeted release date.
“We feel that from a timing perspective [introducing the spec] this year or early next year would not be the smartest move. Customers are wrestling with Y2K problems and Windows 2000 adoption — they don’t need more headaches,” he said.
“Also, we don’t actually feel that there is a fundamental problem in the next year or two that would cause us to want to roll out the I/O system (sooner).”
However, Bill Kircos, a media-relations manager at Intel Corp., in Hillsboro, Ore., said he was surprised to learn of Future I/O’s late release date. He said the industry cannot wait that long for a new specification.
“It’s Intel’s opinion that this should be done fairly quickly. We are already seeing amazing amounts of e-commerce and bandwidth constraints on the Internet today. And that’s going to keep exploding.”
More information on Future I/O will be announced at a conference in Monterey, Calif., on Feb. 12.