Journalists mistrust vendors when they posit platitudes like "Our primary goal is to serve customer needs." Too often, they are the self-serving verbal trappings tacked onto a new product introduction or the explanation as to why another product is late in delivery. With my unabashed cynicism intact, however, I'm encouraged by similar comments from a vendor consortium dedicated to replacing the tried-and-true PCI bus.
Despite its marketing-inspired name, Infiniband is the heir to a long line of system buses - ISA, EISA, Sbus, MCA and, of course, PCI - that have plagued IT operations for decades. Infiniband promises more performance, easier management and increased distance between system components and processors. Data center managers will love it because it will make upgrades, backups, troubleshooting, management and maintenance of servers and attendant hardware a breeze compared with today.
Infiniband has more than 180 backers. It's the result of a merger last summer of two competing technology development efforts - Future I/O and NGIO. Each had its powerful industry supporters, and many observers thought they would ultimately fight it out.
But in a rare stroke of sanity, fierce competitors put aside their posturing in favor of a single, open I/O architecture. In the age of the Internet, vendors reasoned, it made no sense to consciously design incompatibilities into a fundamental IT building block - I/O.
The Infiniband Trade Association (ITA) says it will release its final specification by October. Maybe six or nine months later, low-level components will begin to appear; about 12 to 18 months hence, Infiniband systems will hit the market.
The ITA claims that it needs the time because rigorous compliance procedures are necessary, both in the design of the specification and for real-world intervendor product testing. The ITA vendors say this is because "our primary goal is to meet user needs."
This time, I choose to believe the platitude because user needs and vendor needs are completely in sync. Users building complex data centers will exclude systems with known incompatibilities, meaning a lot of vendors will lose a lot of business. And none of them want their systems ostracized. In that light, the ITA's work is simply calculated vendor self-interest.
It's the kind of self-interest we need to see more often.
Mark Hall is Computerworld's West Coast editor. You can contact him at firstname.lastname@example.org.