Using one dead CPU to emulate another. Let's face it HP you are just a reseller of other peoples IP now. Live with it.
Hewlett-Packard is getting serious about getting customers to move off those vintage HP 9000 boxes, based on its final generation of home-grown PA-RISC processors, and onto shiny new Itanium-based Integrity blade servers. With the launch of HP-UX 11i v3 Update 7 today, HP's flagship operating system will be able to run compiled …
".....Using one dead CPU to emulate another...." Who said anything about using or emulating SPARC?
So, how do you class a CPU as "dead"? If it's by roadmaps then IBM is in trouble - after Power7, the public Power roadmap has nothing but a vague entry of "future" for Power8, whereas Intel have commmitted to another two generations of Itanium using the current socket (yes, that means pin-compatible upgrades for those buying Tukzilla servers now). If you mean "dead" as in showing a historic tendency to DOA chips then you'll be menaing the Sun future high-end CPU that's supposedly arriving in 2015 (and is as likely to arrive on time as SPARC V and Rock). You can't mean Itanium as it seems to be the liveliest of the trio.
So where does PA RISC fit into this analysis, the original dead chip?
And Intel has killed chips it committed to in the past. HP is practically the only vendor left in the Itanic market giving it more than a token effort, everyone else has moved everything to Power or x64.
Actually, Itanium sales have grown. What you may mean is that the UNIX server market (i.e., Power, SPARC, Alpha, PA-RISC and Itanium) has shrunk and continues to shrink as cheaper and ever more capable Lintel/Wintel kit eats into it. Seeing as hp is the largest x64 vendor, hp isn't crying as much as IBM (gradually pulling out of x64 and offloading it to Lenovo) or Snoreacle (third tier x64 vendor?). Instead, hp has looked at ways to use commodity components from the ProLiant range to make their UNIX offering (Itanium) cheaper (which also means hp are leading on pushing trad UNIX RAS features down into the x64 kit), so they can beat IBM and Snoreacle on price as well as performance (countdown to very predictable Jesper IBM benchmark squealing in three... two... one.....).
As reagrds PA-RISC, there's still plenty of it out there, some hp9000 servers having had their lifecycles extended twice due to dips in the economy. We finally retired our last big hp9000 servers last year, offloading them to a broker as it was a better value than the hp trade-in (we had a choice of nice discount off the new kit or a trade-in, so we took the dicsount and went to the broker). The broker was happy to take them, he said he had plenty of customers looking for cheap hp9000 spares (he even had a hosting company in Denmark that was looking for a complete PA-RISC SD32 - any ideas, Jesper?).
".....everyone else has moved everything to Power or x64." Oh, and which vendor do you think has moved to Power? None. All the movement is to x64, not Power. In fact, if you want to include Apple, then IBM has lost a chip customer to x64, not gained any.
"countdown to very predictable Jesper IBM benchmark squealing in three... two... one.....)"
Zero.. Been busy got a new job, moved from operational to strategic. Actually what I found out on some of the work I've done internally to cut cost, is that we make very very little money on x86 offerings. Their short lifespan, combined with the cost of setting them up and pulling them down. Also the software costs per capacity unit is terrible.
The ones we have the best client satisfaction, and we also make a good profit on, as we can leverage the the fact that we have a lot of smart people, for example I think we have one of the best teams of Solaris sysadmins I've ever seen, is UNIX. So we like UNIX. Clients are happy, we are happy.
But we don't like the small UNIX machines, has to have a critical size, and then use a shitload of virtualization. Cheapest UNIX box in TCO, for us, is still the POWER 770 if used right. Now if the client will be content with containers on Solaris, then we can also make a bit of a sweet deal there, although the HW kind of sucks. But let's see how "T3+" turns out. If it has decent single threaded performance then that will be a good thing for our Solaris offering.
"(he even had a hosting company in Denmark that was looking for a complete PA-RISC SD32 - any ideas, Jesper?)."
Actually that could be us, we have no PA SD's on the 'Spare list '. We do on the other hand have quite a lot of Large Last generation SUN boxes like E25K's and older. Even an old Starfire. They need to go to a broker. Otherwise it's mostly massive amounts of HP/IBM/SUN/DELL blades and 1-4U servers.(And I was actually asked by a client team about the performance of the PA 8900 versus Tukwila, and what it would cost to do a migration project from PA->Itanium)
I was also asked to give my 10 cents on the new SD2. Lets just say that the our client architect was quite choked that it was based on a reengineered c-7000 BladeSystem chassis. Not that I was trying to put it down or anything, but I'll put it in the same class as the power 7[7/8]0, which in technology it actually has quite a few similarities with. So we'll have the reverse story now, IBM used to peddle the 570's against SD's, and now HP will peddle SD-2's against 795.
But IMHO it's not in the power 795 class. Sorry.
Good thing about the SD2 is that is much cheaper in TCA than the old SD, which is good. But.. still when we do proposals POWER normally just comes out on top and with POWER7 it's really a nobrainer, HP needs to get their virtualization layer up to speed. Or at least what their sales teams are ready to do with it.
But with regards to the Whole x86 versus the Big UNIX iron. then our UNIX servers normally, looking at the serial numbers, can have a life of at least 5-10 years, going through several generations of upgrades. One of the new POWER 770's in our 'cloud' actually started out as a 1.9GHz p570 back in late 2004, and had almost 3 years of life as a POWER6 machine. And we also have HP UNIX gear that I can say the same about. But x86 is a buy and throw away thing. And the TCO is just terrible, cause client teams are just interested in TCA. They push the cost in front of them, trying to push it to the guy that is going to replace them, when the cost hits them.
"....Why would you want to?...." Well, if you're an old Sun customer you might be wanting to get onto a platform that actually has a future. Currently, that's Intel (Itanium or x64) or Power (sorry, AMD, but unless you really step up I can only see you dwindling into a niche). As regards which is an easier port - to Slowaris or Linux on x64, AIX on Power or hp-ux on Itanium - I'm sure there is plenty of differing views. Seeing as hp is the number one x64 vendor, the likelyhood is they win the hardware deal on the x64 options. On the two UNIX options, I suppose it comes down to preference and experience. As in if you have had the experience of being bent over a desk and rogered senseless by IBM Global Services (currently IBM's way of squeezing as much cash out of a migration), then you may prefer the idea of a toolkit that automates the majority of the tasks for you. And if you're already an IBM customer and suffered the results of many an IBM sizing exercise ("We say we only need x number of cores for your datawarehousing app because our TPC benchmark says so") will be quite keen to escape the clutches of Armonk!
/SP&L, only twice as hard!
There was one bit that made me chuckle:
".....and with a new set of wizards, a simple cluster can be set up in under a minute...."
Hmmmmm. I assume that's if you have done all the prep work and have all your storage and networking ready to go, and not from standalone two hp-ux boxes. If I can dicsount the prep time, then I can have an SG cluster up in minutes with the old release, it's the prep time that matters.
GREAT QUOTE "The Itanium was intended to storm the world, knocking out all other processors (including the x86, Sparc, Power, and MIPS). But thanks to Intel's ineptitude with Itanium, Advanced Micro Devices' genius (and perhaps luck) in getting the Opteron x64 processors out exactly when Intel was weakest in 2003, and Intel's cloning of the Opterons with its Core-based Xeon servers, the Itanium has now been relegated to more or less the same position as the PA-RISC chip was at HP: running HP-UX applications, with a smattering of OpenVMS, Windows, Linux and NonStop clusters."
next time add "But, Micrososft and RedHat have subsequently dropped future support"
".....next time add "But, Micrososft and RedHat have subsequently dropped future support" And then you can add that current instances are still fully supported, and that users that want to start new instances with the current fully-supported releases of Windows (so when is IBM going to get any version of Windows Server supported on Power?) or RHEL (or even both in the same frame with npars - so when is IBM going to get that hardware islotaion capabitlity?) can do so and get new support contracts. After all, we wouldn't want anyone getting the wrong idea, would we?
Biting the hand that feeds IT © 1998–2020