HP gears down NonStops for midrange, emerging markets Hewlett-Packard's NonStop servers tolerate faults even though new CEO Meg Whitman and Wall Street, which is breathing down her neck, can't. And now, in a hope to offset declines on its HP-UX Itanium server business and get its NonStop Integrity machines into emerging markets, HP has trimmed down the configurations and prices on …
Clustering offers High Availability, which gives zero-downtime for some faults, and rapid recovery for others. Fault Tolerance (as done by Tandem, IMP, etc.) is a very different ballgame, where faults cause no downtime at all. It usually required duplicated hardware across the whole system, not just things like RAID disks and doubled networks. It's expensive because you're paying for 2x the hardware to get the performance of 1x. Moving an FT architecture to a different hardware platform is likley to be a little more difficult than just porting an OS. RAC is in no way comparable to an FT system, it is scalable HA.
Both HA and FT have their place, but few businesses outside the Telco space require full Fault Tolerance, which is why few companies have made money selling it.
Judging by recent operational experience, the telcos don't seem to think they need FT kit any more either.
The City doesn't seem to care about who's got the best backroom technology when what they need is the best motivated liars in the risk management section.
No FT? No comment.
They might not need it, but they still try.
I'm not sure how much these cost, but try adding a few F5's into the mix and see how big the price difference is.
I'm not sure about porting to x86 chips either. The problem is usually cheaper kit has lower manufacturing standards and is therefore more likely to fail. At this end of the market, if you need cheaper materials to survive, you're probably dead already - go for quality every time.
HP trimming down on the spec for “entry level” is nothing to do with expanding the base for Tandem NonStop systems, but all about providing development & test servers to try and get new applications written for existing MPP systems.. that will feed into bigger system upgrades.
NonStop is no spring-chicken, the answer is not to port the Kernel to yet another processor but to port the TMF & messaging system to a different OS (think CICS over Encina).. and integrate with platform management & infrastructure virtualisation.. NonStop on OpenStack could provide much needed brand differentiation for the HP/EDS cloud offering.
They might have retired/”best-shored™" all their engineering expertise in Fault-Tolerant TP.. but “NonStop Cloud” is just the conceptualisation that the HP brand needs to transform the vision of its core mission statement/value proposition.
I always laugh when TPM has to write something nice about any hp product as he always starts off piling on the negatives before he can get round to actually putting anything positive into the piece. Why does hp HAVE to port NonStop to x86? I don't see TPM insisting that IBM just HAS to port z/OS to x86. Oh, it's because NonStop is a competitor to IBM mainframe! D'uh!
Why sell machines with one or more cores actually turned off?
Now if the truth was that yields on Itanium are horrific, and there are lots of parts with only one, two or three working cores, that would be a truth the public could accept - but perhaps Intel wouldn't want it broadcast?
If you can't do an upgrade with a jumper, as per early IBM mainframes, perhaps Itanium yields really are the reason why HP wants to migrate to Xeon.
"....which means it has two revs left at most until end of life...." There are two more Itanium CPUs currently detailed on the roadmap, Poulsen and Kittson, and no-one has said with any finality that those are the final Itanium CPUs, definitely not Intel or hp. Even if they were, they would still offer at least fifteen more years of life. And then you have to consider that should hp make a decision to port in ten years time, they may be instead porting to a Xeon-Itanium hybrid with a different feature set, or to ARM (unlikely maybe, but including the option will keep the ARM trolls happy). Either way, hp and their NonStop customers have plenty of time before they have to worry. That is longer than the zOS roadmap can offer given that Power has a roadmap of one more CPU. Instead of just regurgitating the IBM FUD, maybe you should be asking IBM why they aren't planning a port of zOS to x86 by now.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
