Chipzilla sits on its Tukwila Chip maker Intel has quietly copped to another delay with its "Tukwila" quad-core Itanium processors. The company is revamping the quad-core Tukwila processor design, which will push deliveries of the chip out to the middle of this year. And that means system shipments will be even later. This six-month hold-up is by no means …
Tell HP to go packing. Montvale is really just Moncetito with 60 more Mhz. You can't even use the faster front side bus because HP never updated their chipset.
Don't buy Montvale with a "free" upgrade to Tukwila. The Tukwila systems are a complete box swap. Everything changes. The "free" upgrade is after HP's fiscal year end so it's not this year. The swap is core for core so you get less boards/chips/etc.. The other kicker is HP is trying to sell the 18MB chips which have 25% of the cache disabled (defective) to try to make the "free" upgrade easier to justify.
There is never a "free" upgrade. The delay and migration costs do not justify staying with HP-UX. The real reason they are waiting for DDR3 is because Nehalem makes Itanium look so bad.....oops! So when is HP-UX going to be available on Xeon? I hear its already ported.
Cheers,
Matt B.
"....Tell HP to go packing. Montvale is really just Moncetito with 60 more Mhz. You can't even use the faster front side bus because HP never updated their chipset....." Actualy you'll find (if you knew where to look) that the hp sx2000 and zx2 chipsets had been designed with room for stretch built in, which is why you get a 10% performance boost for that extra 60MHz. That's beacuse - whilst not a complete revamp of the chip - Montvale did have a number of tweaks other than just the 60MHz clock jump which made it a faster chip than Montecito.
"....The Tukwila systems are a complete box swap...." Hehe, I still have the Sun FUD slide saying the same about the change from Madison to Montecito, and they were just as wrong then. Yes, the motherboards have to change, but the rest of the systems are likley to stay the same for the meantime. What I'm looking forward to are the Tukzilla blades which should slot straight into our existing c-class enclosures. Much better than he rip-and-replace of updating off UltraSPARC, where your only choice is complete box-swap and pain or complete box-swap and lots of pain.
"....The delay and migration costs do not justify staying with HP-UX...." No, but the ease of migration (exactly the same OS with exactly the same binaries, no need to recompile anything) is a very good reason to stay with hp-ux, especially compared to trying to get off SPARC Slowaris onto any of the new Sun chips (new binaries or recompiles all the way, except for the SPARC64 systems, but then that's because those are Fujitsu not Sun chips).
"....So when is HP-UX going to be available on Xeon? I hear its already ported...." Well you need to go clear your ears out. Either that or stop listening to those voices in your head. Of course, Sunshiners can take hope in the fact that even Sun saw Slowaris on SPARC was doomed years ago, hence the effort to try and get Slowaris x86 back out of the graveyard. Sun try and tell their customers they're still a hradware company, but it's obvious they have no faith in the chances of their server designs from the amount of time they spend begging IBM, Dell and hp to support Slowaris x86 on their much more successful Xeon servers. If Ponytail has such faith in Niagara and Rock as worldbeaters, why does he keep begging hp to support Slowaris on ProLiant?
Of course, Transitive do give current Sun customers a lifeline that when Rock fails to arrive they can port off SPARC altogether onto Power, Itanium or Xeon. If Sun really wants a secure path, they can dig out the old versions of Slowaris they originally crafted for Itanium before they made the mistake of carrying on with UltraSPARC. Seeing as I have benched Oracle on Slowaris 9 on Transitive on Red Hat on an Itanium Superdome as faster than Oracle on Slowaris on an M9000 I'm pretty confidant even an old Montecito could run Slowaris apps faster than Rock ever will - if Rock ever arrives.
So, while Intel make the relatively simple changes to the integrated memory controllers on Tukzilla, Sun are still scrabbling round in the lab, desperately trying to find the holes in scout-threads and transcational memory for Rock. I suspect the truth is Tukzilla will only have to worry about competing with Power6+/7.
"....Tell HP to go packing. Montvale is really just Moncetito with 60 more Mhz. You can't even use the faster front side bus because HP never updated their chipset....." Actualy you'll find (if you knew where to look) that the hp sx2000 and zx2 chipsets had been designed with room for stretch built in, which is why you get a 10% performance boost for that extra 60MHz. That's beacuse - whilst not a complete revamp of the chip - Montvale did have a number of tweaks other than just the 60MHz clock jump which made it a faster chip than Montecito.
==>The faster FSB is not supported and ok 10% better ...BFD
"....The Tukwila systems are a complete box swap...." Hehe, I still have the Sun FUD slide saying the same about the change from Madison to Montecito, and they were just as wrong then. Yes, the motherboards have to change, but the rest of the systems are likley to stay the same for the meantime. What I'm looking forward to are the Tukzilla blades which should slot straight into our existing c-class enclosures. Much better than he rip-and-replace of updating off UltraSPARC, where your only choice is complete box-swap and pain or complete box-swap and lots of pain.
====> there is no in-chassis upgrade. Windjammer is a box swap. I hear you guys are actually going to copy IBM's 570....every think about something inventive
"....The delay and migration costs do not justify staying with HP-UX...." No, but the ease of migration (exactly the same OS with exactly the same binaries, no need to recompile anything) is a very good reason to stay with hp-ux, especially compared to trying to get off SPARC Slowaris onto any of the new Sun chips (new binaries or recompiles all the way, except for the SPARC64 systems, but then that's because those are Fujitsu not Sun chips).
===>the m-class is a joint venture, plus since we own the O/S fujitsu cannot do anything without begging for a license renewal.
"....So when is HP-UX going to be available on Xeon? I hear its already ported...." Well you need to go clear your ears out. Either that or stop listening to those voices in your head. Of course, Sunshiners can take hope in the fact that even Sun saw Slowaris on SPARC was doomed years ago, hence the effort to try and get Slowaris x86 back out of the graveyard. Sun try and tell their customers they're still a hradware company, but it's obvious they have no faith in the chances of their server designs from the amount of time they spend begging IBM, Dell and hp to support Slowaris x86 on their much more successful Xeon servers. If Ponytail has such faith in Niagara and Rock as worldbeaters, why does he keep begging hp to support Slowaris on ProLiant?
====> looks like they were correct that HP-UX has been ported to Xeon....I don't think HP is stupid enough not too
Of course, Transitive do give current Sun customers a lifeline that when Rock fails to arrive they can port off SPARC altogether onto Power, Itanium or Xeon. If Sun really wants a secure path, they can dig out the old versions of Slowaris they originally crafted for Itanium before they made the mistake of carrying on with UltraSPARC. Seeing as I have benched Oracle on Slowaris 9 on Transitive on Red Hat on an Itanium Superdome as faster than Oracle on Slowaris on an M9000 I'm pretty confidant even an old Montecito could run Slowaris apps faster than Rock ever will - if Rock ever arrives.
====> does anyone use transitive for Solaris on Itanium? Forrest Gump.....you have too much access to HP and Sun equipment and time to pretend you are with a biz partner.
So, while Intel make the relatively simple changes to the integrated memory controllers on Tukzilla, Sun are still scrabbling round in the lab, desperately trying to find the holes in scout-threads and transcational memory for Rock. I suspect the truth is Tukzilla will only have to worry about competing with Power6+/7.
=====> Simple? The chip is going to be another 6 months later and systems will not show up until after HP's fiscal year end. Hello? Rock will come out in October along with NiagaraIII
jeez - leave it with the extreme opinions will you guys, this is IT, not religion you know!
- no doubt for those wishing to use a "commercial UNIX", Tukwila will be a solid choice, in the same way Power6 and SPARC64 are. Rock & Niagra will probably have their place as well if Sun can resolve their curreent financial difficulties.
- Did we all forget that competition is a *good* thing? Quit arguing the toss about this sort of thing, or imagine a world with only 2 or 4 socket x86 systems running Linux & Windows - now thats something to be scared of!
- as for porting HP-UX to x86 - well I think HP are intelligent enough to notice the 800lb Linux goirilla in the corner there! I think you'll find that HP will concentrate their efforts in the x86 space on bringing some of the greate Mission Critical fearures and functions of commercial UNIX systems to Linux (with add-on software or in partnership with RH/SUSE - not their own distro), rather than worrying about bringing A N Other UNIX to x86.
- chip delays happen! good grief, ROCK was delayed, Power6 was delayed, AMD delayed some of their systems last year. This just happens - no doubt Intel looked at the economic cimate as well and decided that a 6 month delay wouldn't effect revenues too much anyway (who's buying anything at the moment?)
so there's just as much reason to use HP-UX as AIX as Solaris - these products have their places and will continue to. The market for them isn't going to grow, which is why you see this sort of crazed fighting over the scraps, but neither are any of them going to go away in a hurry.
So HP are going to copy the p570 design are they?
SO you mean they're gonna sell people on the modular design of 4 socket boxes cabled togther, then after you buy a 2-box 8 socket system, they tell you you can't break it up into 2 1-box 4 socket systems?
Some modular solution... I don't think HP are that stupid.
You don't need much nouce to figure out where HP is headed - when they say "blade everything", they mean it.
"Ahhh, isn't that cute? HP fanboy Matt is getting defensive for his favorite company...." Wrong again! I wouldn't want you Sunshiners taling any joy in more fanciful thinking, you have enough of that with your Rock delusions.
"....Oh yeah, Tukwila is already more than 2 years late, so I doubt that DDR3 will save it...." Save it? Strange, it looks like it will be going into servers from a company actually growing in the high-end UNIX arena which provides big margins, doesn't sound like it needs much saving. By the way, how many sockets will Rock top out at (if it ever gets out)? Say b'bye to the high-end, Sunshiner!
"....That's all..." No, please continue, there's another couple of minutes of ads on Paramount Comedy to get through yet and you provide such good if unintentionally comic material
No no. That's OK Matty Boy. We understand how seeing your favorite company bet on the wrong technology is giving you much pain. We all notice how you include the term "high-end" Unix when you brag about HP... The fact is that HP is losing overall market share in Unix and the only thing keeping them growing are the old PA-RISC and Alpha captives moving over. Once they are all moved over there will be nothing to artificially keep Itanic afloat. Very sad indeed.
"No no. That's OK Matty Boy. We understand how seeing your favorite company bet on the wrong technology is giving you much pain...." The only pain I'm getting is the split sides from laughing at you Sunshiners so much. Face it, Sun is dead in the corporate, and dying everywhere else. Ponytail says he has loads of interest in cloud, but lets slip that it's all from low-level techies and not the CIOs that have the actual money to spend. If anyone is making a mistake, Ponytail and cloud is just another in a long litany of Sun mistakes.
"....We all notice how you include the term "high-end" Unix when you brag about HP...." Ah, but you neglect to mention that is the exact area Sun used to brag about owning so much. The fact that is Sun cannot survive as is without making the kind of revenues the high-end brings, which is why it's shares are rated as junk. You can bluster and bleat as much as you like, but hp will still be making a profit when Sun has just become an example used to teach buisness students of what happens when you don't diversify and don't understand what customers want.
"....The fact is that HP is losing overall market share in Unix and the only thing keeping them growing are the old PA-RISC and Alpha captives moving over....." Sorry to burst your bubble, but IDC and Gratner both say you're lying. Seeign as the new Integrity kit does more for less, to stand still hp would actually have to sell more. Seeing as they are growing revenue with Integrity and gaining marketshare that means they are not just migrating those old PA-RISC and Alpha customers but also taking share from others, particularly Sun. And I'm sure more people will believe IDC and Gartner than you, no matter how long and loud you squeal.
"....Once they are all moved over there will be nothing to artificially keep Itanic afloat...." Nothing artificial needed, seeing as Integrity makes a profit for both hp and Intel, unlike any of the Sun server ranges which Sun has to prop up by dipping into the reserves. What do you think will happen when all the reserves are gone, moron? I'll try and keep it as simple as possible for you. Sun is a big old bucket with a big hole in the bottom that is the cash going out, and a tiny tap flowing in that is the revenues from all their products. Even should Sun DOUBLE their server sales they would still NOT match the flow out the hole. Sooner or later the bucket will run dry, but before then Southern Asset Management will have broken up Sun and sold off anything they can to get their investment back. Even if Ponytail keeps trying to climb up Obama's backside and asks for a "stimulus" handout, it will just delay the inevitable. Hope you've managed to learn some Linux before then.
Matt,
You still haven't explained why Itanic is more than two years late and the specs are almost half of those promised. One other question is why every Itanic release is late. My guess is that the design of Itanic is overly complex and does not provide for quick design wins. Itanic has no future as anything more than a way for server vendors to get to X64. The shared bus is the first example of why this is true. X64 is OK on the low to mid end, but trying to run a real high-end on the x64 architecture is death.
Also, Matt, the post was talking about Itanium here, not SPARC, though I know you are singly focused on Sun for some reason. Please, please, please, don't start talking about Sun again. No one wants to hear you. Just give us some reasoning for the questions I asked. Don't try to hide from Itanics problems by pointing at Sun's problems.
"You still haven't explained why Itanic is more than two years late and the specs are almost half of those promised..." I don't know which roadmap (if any) you are reading from but I'm guessing it's very old, probably from about the same period as Sun were promising something called UltraSPARC V, and probably long before UltraSPARC IIIi+ disappeared too. Please provide a link to any Intel doc that says anything of the like.
"....One other question is why every Itanic release is late...." Compared to Sun chips they're positively timely. But then most chips are late, the recent Atom release being an exception. But the real exception is how many chips Sun claims are "strategic" and then buries. Now, how has Sun been referring to the three-year-late Rock? Oh yeah - "strategic"!
"....Itanic has no future as anything more than a way for server vendors to get to X64...." Strange it's doing so well in the UNIX high-end then, especially as it is taking share from SPARC Slowaris. Must be a lot of Sun customers real eager to get onto x64 on hp rather than Sun kit (Galaxy really must be that bad).
"....The shared bus is the first example of why this is true....." <Sniff, sniff> Smells like male bovine manure to me! It's just like when hp and IBM started using high-grade components from their x86 servers in their UNIX kit and were thus able to lower the price without compromising reliability. Sun didn't have the same benefits of a large x86 base to gain from the economies of scale from such a move, so Sun started spreading FUD about hp and IBM using "cheap PC parts". Fast forward a few years and Sun has bought in x86 designs in a desperate attempt to slow their downward spiral, using those same "cheap PC parts". The new common socket design for Tukzilla and next-gen Xeon is another smart move as it will allow vendors to share more components between x64 and Itanium ranges, which means they will still be massively cheaper and more powerful than any server Sun can make (or buy badged from Fujitsu). And Sun will still not be able to match hp or even IBM becuase Galaxy has such small marketshare and Sun has so little cash it cannot afford to spend on new server developments, it has to follow where others lead.
"...X64 is OK on the low to mid end, but trying to run a real high-end on the x64 architecture is death...." Strange thing to say seeing as Sun's whole future hangs (by a thread) on Slowaris on x64. All the evidence is there - the abandoning of the high-end, the failure of Niagara to climb out its narrow niche, the lack of date for a successor to SPARC64 VII, the desperation with which Sun have chased hp, IBM and Dell to try and get them to resell Sun support for Slowaris x86 on their x64 server ranges. When IBM announced they were supporting Slowaris on x86, an IBM salegrunt said to me; "That's it, SPARC is dead, if they get on ProLiant as well then they'll be a software business only within five years." What he should have said is that Sun will be dead and gone in ten years because Red Hat and Novell will beat it in every way on x64, just like Power and Itanium have hammered it in the UNIX arena.
"....X64 is OK on the low to mid end, but trying to run a real high-end on the x64 architecture is death...." Which is why hp have Itanium in Integrity (you remember, the CPU that allows you to run multiple datacenter OSs in the same chassis, the one taking so much share from Sun in the high-end?). Which is also why hp are the number one server vendor, because customers know hp can supply integrated solutions from the desktop through to the top of the high-end (ever heard of NonStop?), backed up by a trusted support structure and a range of management tools, storage, etc, etc. Oh, and those printers you Sunshiners hate so much!
"....Also, Matt, the post was talking about Itanium here, not SPARC, though I know you are singly focused on Sun for some reason. Please, please, please, don't start talking about Sun again. No one wants to hear you. Just give us some reasoning for the questions I asked. Don't try to hide from Itanics problems by pointing at Sun's problems...." Ahhhhh, don't cry. You Sunshiners tried to make out that Tukzilla being delayed by six months so it could take advantage of DDR3 (by the way, how many times faster will that be than whatever memory Rock gets, should it ever get out, and cheaper?), so I merely used the example of Sun's awful history, poor chip record and dire financial position to show you that Tukzilla really isn't in any trouble at all. After all, if you look at the public roadmaps, Intel's has three coming generations of Itanium (Tukwila, Poulsen and Kitson) with dates on the roadmap, whereas IBM only has Power7 (no set date, just marked "Future") and nothing to follow, Fujitsu has no date for SPARC64 VIII and nothing to follow it, and Sun's roadmap for developments after Rock and T2 have no dates and should really be filed under "fiction".
"three coming generations of Itanium (Tukwila, Poulsen and Kitson) with dates on the roadmap"
You're joking right? Intel has code names and Poulsen is supposed to be out at the same time that Tukwila is now coming out. There are no specifics, just code names. You call that a roadmap?
Also, no one's concerned about the six month delay, it's the two year delay that Tukwila had before that. Rock has always been expected in 2008, so the one year delay is nothing compared to the constant Itanium delays. You said you wanted to see a link to the 2007 date for Tukwila? Well, a real quick search on El Reg shows:
http://www.theregister.co.uk/2003/03/26/tanglewood_the_next_itanium_revealed1/
http://www.theregister.co.uk/2005/02/24/intel_nicks_tukwila/
http://www.theregister.co.uk/2003/07/10/intels_tanglewood_pumped_full/
Hmmm. I wonder which is faster with better technology? Huh? Sun M9000 or Superdome? Let's see... Sun beats HP on SAP-SD, TPC-H, SPECCPU2006, SPECjbb2005, SPECOMP, LINPACK, SPECjbb2005, and will be able to support future CPU's (SPARC64 VII and VII+) in the current chassis, while HP will require a forklift upgrade with Tukwila. As a matter of fact the M9000 can support different CPU's in the same domain running at native speeds, while HP does not allow this.
RE: Re: another Matt B. Rant
"....You're joking right? Intel has code names and Poulsen is supposed to be out at the same time that Tukwila is now coming out. There are no specifics, just code names. You call that a roadmap?...." Once again, those outside Sunshinerville are better informed. The Intel roadmap I have is NDA, I assume if you were trusted by hp or Intel or just an important enough customer you'd have a copy. Anyway, it not only has Poulsen and Kitson but comments on the technologies to be expected with each. Now that's a roadmap, much more so than any available from Sun or Fujitsu.
"....You said you wanted to see a link to the 2007 date for Tukwila? Well, a real quick search on El Reg shows:...." No, I asked you to provide a link to the material you claim said Tukzilla will arrive with "specs are almosts half those promised", but then I suppose it was too much to ask for you to be actually able to back up your wild ramblings with some form of factual material. As with many of Sun's products, your FUD is just so much hot air.
RE: M9000 vs. Superdome
Every benchmark you mentioned off the Sun website are against a 2006 Superdome, and even then you had to fully-stack the M9000 with the latest chips (and double the licences costs) to do the job and under Sun's supervision - yeah, I bet they know how to tune hp-ux - not! My own practical experience from doing proof-of-concepts (we do all our major deals on try-before-you-buy, so the vendors have to back up their sales stories with actual performance before they get the cash) is that Superdome not only out-perfroms the M9000 (Oracle and SAP) but also does so massively more cheaply. If you want to argue the point, go ask SAP which vendor has the most SAP installs and the most SAP on UNIX installs (big clue - it ain't Sun and by a large margin). There is a good reason hp is eating up Sun's high-end business, and it because the customers find that kit like Superdome outperforms Sun's offerings, at a lower price, and with better reliability, features and support. Rant all you like, the market figures speak for themselves - Sun is dead in the high-end.
"....while HP will require a forklift upgrade with Tukwila...." Well, seeing as the current Superdome frame has supported Itanium and PA-RISC CPUs since 2001 (a lot longer than any Sun server has supported any SPARC range) and seeing as new cell-boards are required for Tukzilla, it seems like a good time to refresh the design and throw in some new bits. Of course, those customers hp have sold Suprdomes to over the years have known for a long time the change was coming, so your bleating is again just the same old and very out-dated Sun FUD. Can't come up with something new?
"....As a matter of fact the M9000 can support different CPU's in the same domain running at native speeds, while HP does not allow this...." No real need as hp can support different Itanium speeds in different hardware partitions, and can even support mixed frames with PA-RISC CPUs and Itaniums in different hardware partitions. Can Sun mix SPARC64 and Niagara in the same chassis, or UltraSPARC and Niagara? No, they can't. Because Sun's diefferent chips are so incompatible you not only need different systems you need different versions of the OS and different binaries.
Tukzilla will arrive this year, with a performance boost over the curent Itaniums, and hp will have servers ready to ship. Compare this to Sun, who will still be trying to get Rock working let alone have server designs ready for it. It really has become a two-horse race between hp and IBM.
"under Sun's supervision"
That's an argument? The rest of your comments are laughable, but this one is just stupid. Are you telling me that you really think that the HP benchmarks are done by people that have nothing to do with HP? You really are gullible. No wonder you buy the HP/Intel lies so easily.
"....Are you telling me that you really think that the HP benchmarks are done by people that have nothing to do with HP?...." Which is why we tend to ignore competitive data from vendors and go with standard comparison benchmarks from TPC and SPEC. We then go to proof-of-concepts where the vendors have to prove their kit does what it says on the tin with real data in our environment - no pass, no cash. We're not gullible and we don't buy anyone's lies, we buy kit that does the job with the least fuss for the lowest overall price. As a result, at the moment we are not buying any Sun, we're predominantly buying hp and some IBM. It has been very revealing that on more than one occaission, when we announced we wanted a shoot-out to compare offerings, hp and IBM stepped up but Sun backed out. If you had a clue about real-world procurement and operation you'd know about the hoops we make the vendors go through to get our cash, but then I don't suppose you see much of the sharp end from where you sit in Sunshinerville.
I couldn't agree more, for two very good reasons. Firstly, the IBM system may actually do the job better for the same or less money. Secondly, if you introduce competition into the purchase you are likely to increase the discount or get extras such as services thrown in for free by one or other of the vendors.
So, apparently Matty is the only "user" doing performance analysis and all of the companies buying Sun are too stupid to do shootouts. A hint for you Matty, Sun sells twice the number of Unix servers that HP does and sells more Unix servers than HP and IBM combined. That must mean that most Unix purchasers are too stupid in Matty's opinion...
Nice Matty. Go back to your HP marketing drawing board...
Oh yeah Matty, one other thing. SPEC and TPC don't do the actual benchmarks on the equipment. Users and vendors do. Apparently you don't understand this stuff as much as you think you do. Apparently, if you're doing SAP, you don't think that SAP-SD is a valid benchmark? You actually think that SPEC or TPC would tell you how your actual load would work better than the actual load that you intend to do? That's very telling... You're probably the type of person that does a copy command from /tmp to a block device and uses that number as your performance comparison. Too funny! Microbenchmarks, you gotta love 'em.
Boring Bill strikes again! I really wonder if Bill lives in a parallel reality or if their is some logic-defeating forcefield surrounding all Sun facilities that stops him realising what twaddle he writes.
"So, apparently Matty is the only "user" doing performance analysis and all of the companies buying Sun are too stupid to do shootouts...." Nope, there are plenty of customers just going straight to Sun. What you forgot to look at was why - because they made poor strategic decisions in the past and backed UltraSPARC, kept those old Slowaris systems as long as they can in the desperately vain hope Sun would get Rock out on time, and now their upgrade options are so limited they really have a choice of pain, lots of pain or just give up and start from scratch. Unless they want to start from scratch and port to another vendor - unlikely in the current downturn - they have to go for Sun, but in the longterm they will be looking for other options as they are business people, and they won't like betting their livelihoods on the poor odds of backing Sun. The proof is in Sun's decline in the high-end, which shows that businesses just don't believe they can trust Sun with their business critical requirements.
"....Sun sells twice the number of Unix servers that HP does and sells more Unix servers than HP and IBM combined....." Yes, we've already been over your Sun figures and had a good laugh. Sun are selling lots of small boxes at small to no margin, whereas hp and IBM make more revenue per system and pull through more services and support revenue. What you are deliberately ignoring is hp and IBM each make more money out of Power and Itanium than Sun do out of all their server lines put together - SPARC, Niagara and Galaxy. And seeing as Sun doesn't make enough money to make a profit, this means Sun will have to stop doing servers pretty soon. It's simple maths, which is why Sun's market cap is brown and smelly.
"....SPEC and TPC don't do the actual benchmarks on the equipment. Users and vendors do...." Yes, the point being they are testing their own kit, not deliberately misconfiguring competitor's kit to make theirs look better. And they have to disclose how they did it, so you can see if it's with a stack that is unrealistic and won't benefit your environment or application. Compare this to Sun's copmpetitive "benchmarking" where they hide all the real info away so you can't see how poor a comparison it really is. If you made a decision of just Sun's "benchmarking" then I wouldn't call you stupid, just incompetent.
"....Apparently you don't understand this stuff as much as you think you do...." No, you just leapt to the wrong conclusion due to poor comprehension of what I wrote. But then I'm not surprised your preconceptions and fear make it hard for you to take things in.
"....Apparently, if you're doing SAP, you don't think that SAP-SD is a valid benchmark?...." I think getting the kit onsite, building it how you want it with your own images, patches, software stack and SAN, and then benching it using copies of live data, is more valid than any vendor's benchmark. That way you know for sure before you have to pay that it is going to do the job.
"....That's very telling..." It's very telling that you get upset at the idea of competitive shootouts and you'd rather sprout Sun's "benchmarks" instead.
"....You're probably the type of person that does a copy command from /tmp to a block device and uses that number as your performance comparison....." Like I said, I do real benchmarking in real environments with real data and then use that to pass recommendations to the board. I suspect you have never done a real bench in your life, wouldn't even know the way to your boardroom, and couldn't formulate a recommendation without getting it written for you by the Sun salesgrunt.
"....Too funny! Microbenchmarks, you gotta love 'em." Sunshiners, so funny, but you just end up pitying them.
Matt, you lack logic and consistency. Because of this, I can't help but concentrate on others that actually make sense more than 50% of the time and ignore your blathering. You are the one that is always spouting the fact that Sun has no benchmarks and then when everyone, not just me, quotes them to you, you ignore them... You can't have it both ways Matt. I've done competitive shootouts between HP, Sun and IBM. The fact is that IBM comes out on top on Oracle because of the licensing issue, Sun comes out on top for everything else, and HP never comes out on top except in some very specific microbenchmarks. Of course there are some exceptions, but in general this is how it goes. Sun scales from low-end to high-end while IBM and HP have chosen to only concentrate on the high-end. This makes HP way too expensive to even consider. IBM and Sun however are the way to go. Why anyone would ever want to deal with HP-UX is beyond me. Even HP has given up on advancing this incompetent OS and are trying with all their might to push everyone to Linux, which has it's place but does not fit all needs at the moment.
You are an oaf and a liar sir.
"....You are the one that is always spouting the fact that Sun has no benchmarks and then when everyone, not just me, quotes them to you, you ignore them...." Nope, I said show me these wonder benchmarks you keep going on about, and you couldn't provide industry standard ones like TPC or SPEC without admitting the Sun performance was pants, you just kept pulling out the Sun sales guide and pushing Sun's "compatitive benchmarks". When I quote real benchamrk figure done in shoot-outs you just flip out.
"....The fact is that IBM comes out on top on Oracle because of the licensing issue, Sun comes out on top for everything else, and HP never comes out on top except in some very specific microbenchmarks...." Well, seeing as hp Integrity is the king in the high-end, it would seem those "very specific microbenchmarks" happen to coincide very neatly with what real corporate customers think they need for business critical applications. Come on, at least try and see past your own FUD and admit that if hp-ux is king in high-end UNIX then it has to be by customer choice, which means hp are doing something better than Sun or IBM. Go on, just try.
"...Sun scales from low-end to high-end..." Sun scales very expensively in a disjointed fashion across several incompatible server ranges with different binaries from an over-priced low-end to an unwanted high-end. Sun's only hope is getting other vendors to support Slowaris on x64 kit.
"....while IBM and HP have chosen to only concentrate on the high-end....." IBM and hp have true scale, in hp's case from the rx2660 right up to Superdome, and with Integrity blades in the most popular blades chassis on the market. Or did you forget again that hp is the number one server vendor? No, I'm sure you didn't forget, you just don't want anyone else to remember.
"....This makes HP way too expensive to even consider...." Proof positive you don't work in the industry, otherwise you'd realise the high-end is one of the most cost-competitive arenas around. Sure, the pull-through of services and support are bigger, but costs are a major issue, and if hp-ux was so expensive as you insist then it simply wouldn't be number one. But it is, so obviously you're just talking more male bovine manure. Face it, when Tukzila arrives it's going to be running the number one UNIX from the number one server vendor.
"....Even HP has given up on advancing this incompetent OS and are trying with all their might to push everyone to Linux...." Strange then it's the number one UNIX in the high-end. I guess all those highly-paid CTOs and CIOs are just to stupid to ask for a roadmap or any commitments from hp? Yeah, right! There is a public roadmap for hp-ux with plenty of development targets listed, and hp are making a profit from hp-ux, unlike Sun which are still making a loss on Slowaris. Sun are so desperate they're trying to give Slowaris away for free in order to get some support contracts, but it's not working. But becasue hp-ux is still valued by users, especially in the high-end, hp is still making plenty of money from support and licences. You really should try researching a bit more before going on your bitter little FUD trips. It is so obvious you just can't stand the fact hp is making a profit from Itanium and hp-ux, and Sun just can't make a profit fullstop.
"....You are an oaf and a liar sir." Prove me a liar then. Oh, you can't. In fact, all you've proven is that Sunshiner FUD is still a tragic comedy. So why don't you crawl back to Guillemont Park and hide under your desk until it's all over and Sun is broken up and sold off, then go get yourself a McJob you're more suited to.
Too funny. Matt bases his argument on the fact that HP is the number one vendor, so they must be right. It actually has some merit, but not in the way that Matt thinks. The next couple of years will be very interesting.
I think that HP will slowly die on the high end. IBM will overtake HP in servers again. And who knows what Sun will do, but my guess is that they will continue to be the best Unix.
Dear plebs,
Once again it pains Me that I have to take time out from My Work To Make The World A Better Place to explain to you undeserving cretins that My Acolytes must not be mocked! Bill is one of the Chosen, he has seen The Light and the greatness of My Ponytail. From now on you will accept all his utterrences as though they are My Own! Well, actually they are My Own seeing as he has had the unneccessary ability of objective thought removed to allow him to understand My Plan. Nevertheless, just accept My Word that all Sun servers are infintaley superior, much cheaper and make cinnamon toast, and that Solaris is The One True OS that leads to Salvation! Forget this foolish prattle of benchmarking, shoot-outs and comparisons, they are the talk of The Devils hp and IBM! Now, get on with My Bidding.
Yours Heavenly,
Jonathon I Schwartz.
"....I think that HP will slowly die on the high end. IBM will overtake HP in servers again. And who knows what Sun will do, but my guess is that they will continue to be the best Unix." Whilst an epitome of lucidity compared to Bill's dribbling, would you care to supply any reasoning to support your ideas? With IBM gradually abandoning the low-end x64 to Lenovo, I predict it will be very hard for IBM to match hp's unit figures. And as for Slowaris being the best Unix, whlst you are entitled to your opinion, it is not shared by people like Gartner, whose Magic Quadrants rate hp-ux higher than Slowaris or AIX.
I'm not a big fan of Gartner. Way too many journalists and way way too many so called CxO's rely on what they say. In this particular case I have seen the the Gartner reviews of Sun and HP and they rate Solaris as "Very Positive" and HP-UX as just "Positive". I have not seen the magic quadrant that only discusses OS's... If what you say is true, then it only shows how two faced Gartner is and how little anyone should pay attention to them.
As far as I'm concerned, Solaris is better as I can run it across sub $1,000 boxes all the way up to multi-million dollar boxes. Solaris is better because a quick look on a job board shows 2,334 jobs for Solaris and only 403 for HP-UX (1,233 for AIX by-the-way). Solaris is better because there are a LOT more applications for it - the last numbers I saw show that Solaris for x86 even has more supported applications than HP-UX for Itanium, something like two-to-one. There are a ton of usability issues that I like, but I've run out of time and usability is more of a preference thing.
Matt,
Since you seem to think that Gartner is king. Let's look at what their opinion is on Sun and HP on this metric that you have brought up.
http://www.hp.com/hpinfo/newsroom/awards/hp2354.pdf
http://mediaproducts.gartner.com/reprints/sunmicrosystems/article4/article4.html
So...
Solaris = Strong Positive for Sun
HP-UX = Positive for HP (and that's actually an increase)
Are you just making this stuff up as you go Matt or are you just a liar?
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
