Slashdot Mirror
HP Asks Judge To Enforce Itanium Contract Vs. Oracle
Dupple writes with this quote from a Reuters report:
"Hewlett-Packard Co told a judge on Tuesday that Oracle Corp should be ordered to make its software available on HP's Itanium-based servers for as long as HP sells them. Lawyers for HP and Oracle presented closing arguments in a California state court for the first phase in a bitter lawsuit between the two tech giants. ... Oracle decided to stop developing software for use with Itanium last year, saying Intel made it clear that the chip was nearing the end of its life and was shifting its focus to its x86 microprocessor. But HP said it had an agreement with Oracle that support for Itanium would continue, without which the equipment using the chip would become obsolete. HP said that commitment was affirmed when it settled a lawsuit over Oracle's hiring of ousted HP chief executive Mark Hurd. HP seeks up to $4 billion in damages."
Where I work, we use Alpha, and have investigated moving to Itanium, but with this debacle, we won't make the move, as we use also Oracle. I won't discuss actual numbers, but $20,000 per CPU is in the ballpark of our annual support costs (if you include hardware and software). At some point we will either port or rewrite, and in either case, it won't be an HP platform we land on, but we may still use Oracle for the database. This case has very real implications for HP, and hurts them more than it hurts Oracle.
Because they have contracts with their customers guaranteering them continued software support. And if the main supplier stops software support, those contracts become quite shaky.
The original Itanium was a disaster, the second generation was what the first should have been but wasn't, the third actually looks very respectable. Intel would be stupid to eliminate a product they've actually got functional.
And for high-end use, the Itanium is a genuinely useful CPU. Because the performance of a cluster is a function of the communication delays, very high-end clusters WANT to have very high-end CPUs. You can only do so much with piles of PCs before the inefficiencies due to (a) distance and (b) an inefficient architecture really set in. There's also (c) - a crap instruction set - but the Itanium doesn't help much there because although it is somewhat better, nobody has built a particularly good compiler for it yet. Optimization on the Itanium remains a challenge.
Admittedly, it's not the design I would have chosen - I far prefer many of the design decisions made in the Inmos Transputer and the Intel iWarp, since those were designed specifically for the purpose of clustering and started from that position. I also prefer the elegance of the MIPS64 instruction set over the unnecessary burden of anything Intel has done, but again I'm in the minority. I'd also have threaded compute elements and produced virtual cores, rather than threaded instructions on physical cores, since threading the compute elements would allow you to distribute the heat better, wouldn't prevent you accessing elements that are wholly independent of those in use and would reduce unnecessary swapping. But what do I know, I've only been observing what actually works vs what the customers want for 35 years. Customers are just as stupid as beancounters and pointy-haired bosses.
It's a small world and it smells funny; I'd buy another if it wasn't for the money; Take back what I paid (SoM)
Itanium servers are very profitable for HP and they don't really have any competition in this server space. It is very expensive for Intel as they don't sell enough of them to justify the R&D and support. Intel would like to drop it but for HP. Oracle would like to drop it for support costs. Probably HP wants to keep their profits even though it costs money for everyone else.
Well, there's spam egg sausage and spam, that's not got much spam in it.
HP have been screwing their customers for a while already...
Killing PA-RISC and Alpha, forcing customers onto IA64 which now looks like it will die soon too...
With each processor transition, customers at the very least have to recompile their code or run it under slow emulation...
Sun and IBM snapped up a lot of customers from HP over the IA64 transition, and i fully expect them to do the same when HP finally abandons IA64 and moves its customers onto something else.
Introducing a new, incompatible CPU was never going to work...
Too much code is only available in binary form, and so ran on ia64 very slowly under emulation. Vendors had no incentive to port their code because there were so few users, and users had no incentive to buy into the architecture because there was no software. Who wants to pay 5 times more for a cpu that when saddled with emulation is actually slower than the cheaper cpus?
The hardware was never cheap enough to attract hobbyist developers, so even open source code was often not well tested on them (nor are new risc cpus from other vendors, but old ones can be had cheaply). Had the hardware been cheap, it would have gained a lot more traction in the linux market at least.
http://spamdecoy.net - free throwaway anonymous email - avoid spam!
It seems to me that HP would be better off sinking this money into contributions to PostgreSQL / EnterpriseDB; it already offers a ton of Oracle compatibility, and runs on HP-UX: http://enterprisedb.com/products-services-training/products/postgres-plus-advanced-server/advanced-server-oracle-features
HP already ported it to x86 as internal experiment, it can run there
http://www.theregister.co.uk/2012/06/08/hp_ux_on_x86_project_kinetic/
One of our customer bought some Itanium servers from HP, 3-year old unused hardware, they must have done a deal they couldn't refuse. Then the Oracle wouldn't run fast enough and of course it was our problem, us being the software vendor. After expensive CPU upgrades and everything else, it still runs like a dog. For the fraction of the money, they could have got some intel kit and run Linux with 10x performance.