Are Intel and HP beating a dead horse through their continued support of the Itanium processor? Or are detractors simply seeing a glass half empty because the device, while competent, has not lived up to the heady promises of its introduction?
Mixed metaphors aside, it appears that the Itanium is having a harder and harder time justifying itself as a high-end enterprise processor, leaving it very little wiggle room as strictly an HPC solution. In the end, Intel will have to be the one to decide whether the returns justify continued support for the platform.
A major shoe dropped on the Itanium this month when Red Hat announced it will not support the Itanium on RHEL 6, the next major upgrade to its Linux OS. Very few RHEL distributions exist on Itanium machines to begin with, so it appears the company looked at the cost-benefit of continued support and decided the numbers simply did not add up. The company will continue to support the chip on RHEL 5 distributions until 2014, and it is likely that a few OEMs may extend their support for a few years after that.
That leaves Novell as the only major Linux supplier to still back Itanium, for the moment. As the Register points out, Novell has not tipped its hand one way or the other on the Itanium, and could very well extend support as a niche market, the way it does for the System z mainframe. But it seems clear that for most Linux providers, the IBM Power processor will be the high-end platform of choice.
HP, for its part, seems to be holding up its share of the Itanium bargain. It is on the verge of launching a new line of Integrity NonStop servers featuring more capacity, faster processing and support for open-source software like Spring and Apache Axxis2. The package includes new clustering software, NonStop BladeCluster Express 1.2, which scales performance over thousands of Itaniums and even provides WAN support for broadly distributed architectures. It also includes new SOA software and a new SQL release.
The question for HP, though, is whether enterprises will see fit to continue supporting high-end hardware deployments in the age of the cloud. The company's own services and strategy for enabling internal clouds, like the HP Operations Orchestration stack, might cause many enterprises to view high-end Itanium systems as a luxury they can't afford, particularly when they can get suitable performance and scalability through their own clustered platforms or from third-party providers.
At the moment, the enterprise industry is in such a state of flux that neither HP nor Intel can be blamed for hedging their bets. The Itanium represents a lot of R&D dollars, and it would be understandably difficult to pull the plug before the next decade's hardware development patterns are truly set. Perhaps legions of cloud providers will find they can't live without high-end, fault-tolerant hardware even as the typical enterprise draws down its own infrastructure?
Only Intel truly knows if the Itanium has provided a sufficient revenue stream to justify its continued existence, and they've kept that information pretty close to the vest. The biggest mistake they could make now is to throw good money after bad by chasing a dream that never was.