For all the talk about the competition between server platforms, very few machines are in play at any given time. And this is particularly true of big iron, like high-end RISC/Unix servers, proprietary midrange gear, and mainframes. But when times get tough, those venerable back office systems start looking expensive, and that's …
BART - old reference
The BART "win" is old and is actually a PeopleSoft reference. Most PeopleSoft installations are when people move incredibly old self written HR apps to a packaged applications. Why they chose a declining HPUX and itanic anyones guess. Maybe Carly gave them too good of a deal to refuse. As far as the savings.... I am still waiting to see how the $10 Billion Itanium Alliance investment is doing....seems to be a marketing gaff now. My guess is they are up to $1.5B and half way thru the timeline.
You realise that you work for a big company when they've got 1% of the worldwide IBM Mainframe footprint in their datacentres...
I'd never really bothered to think how few IBM mainframes there may be.
Out of frying pan and into fire
If you're getting off a proprietary hardware platform like Z-series mainframe, why would you jump to another proprietary platform like HP_UX or AIX?
Perhaps you've just got used to spending large sums of money and it's too much of a step down from the addition to putting vendors' kids through college.
Especially if you're moving to Linux, my recommendation would be x86 platform.
There are some things that you cannot move, due to application portability, software vendor support, etc, but for the other 95%, get going: this is a purely financial argument and the days of this platform not being reliable enough are so 4 years ago.
Would anyone disagree - and have a pragmatic argument to the contrary?
"this platform not being reliable enough"
Quite. I would be fascinated if anyone could come up with a couple of plausible technical Reliability, Availability, Serviceability features found in mainframe or even "enterprise server" class systems (Solaris boxes, Itanium boxes, Unisys boxes, whatever) which can't also be found in suitable industry standard x86 servers (eg decent Proliants). It strikes me that in the last decade or so, in the vast majority of systems there are two RAS-limiting factors neither of which relate directly to the core hardware: (1) the quality of the OS in terms of core reliability and its available features to support best RAS (eg meaningful event/error log and crash dump analysis facilities) (2) the quality of the hardware and software support for anything other than the most trivial of problems ("have you reseated the SIMMs" doesn't qualify, and nor do suppliers that don't have anyone outside HQ who knows how to use their crash dump tools).
And then there's "Common System Interconnect", aka Quickpath, on the horizon (still). In principle, if it ever arrives for real, the same socket (and thus the same system) can run with either "enterprise class" chips (Itanium) or with "industry standard" chips. What does that do to the overlap and the competition between the RAS features on IA64 and the RAS features on x86-64? Once Quickpath is out, what will be Intel's (and indeed HP's) motivation to carry on investing in IA64 hardware when they'd get a better return on their investment in the x86-64 market (or in HP's case, probably in higher margin ink products).
"Common System Interconnect" is the death of Itanium
Itanium is going bye bye. HP would be doing so much better if they had Xeon systems which scaled beyond 4 sockets. CSI will finally give HP customers scalability with out the nightmare proprietary hpux only itanium.
It is the old story-- the cycle price is soooooooooo low for an i86 vs one of those old dinosaur things, you should switch over.
Doesn't matter a whit if the application you need done once every 30 days now takes 62 days, you are saving gobs of money. Well you would be saving gobs of money, but now you are tied to proprietary HPUX and that pesky support contract... per unit price looks suspiciously similar to the old mainframe price. But the cycles are soooooo cheap... ignore that man behind the curtain, just remember the cycles are sooooooo cheap.... even having to have duplicated servers and more proprietary software to run in parallel for reliability is OK because the cycles are sooooooo cheap....
Is the Reg going soft on HP?
Shoulkdn't we take HP claims about customers switching from mainframes for reduced Total Cost of Ownership (TCO) with a pinch of salt?
Well, the analyst whose TCO report was cited by HP was paid for his work - by HP. Moreover, HP's release makes no mention of market share leadership or gain by HP, signaling they don't think they have the leading share.
Then there are credible redports from third parties, like IDC, which say that System z has nearly doubled its share this decade, while HP and Sun have lost share.
What do you think? And who do you believe?
Shame on the Reg for swallowing this HP PR - where is your normal healthy cynicism?
Is the Reg going soft on HP? (part deux)
Statistics, lies and PR .....
Interestingly enough, earlier this week, Robert Frances Group issued a public statement -- as a result of HP's press release which cited RFG data and claimed that 250 customers moved from mainframes to HP Integrity servers. The statement is available at
RFG noted that its own analysis was flawed -- using apples to oranges comparisons and indicating that HP misused the data and drew inaccurate conclusions. RFG confirmed that they still believe the mainframe is one of the best and most energy efficient platform options and is not suggesting platform abandonment for cost reasons.
I'm not advocating one platform over the next - that's for clients, analysts and journalists to decide for themselves - but they need to retain a healthy cynicism and avoid swallowing HP's PR and dodgy ststistics. Come on Reg, you know you're better than this.