Oracle defies HP and IBM with 47% revenue leap If you were thinking that that acquisition of Sun Microsystems was going to drag Oracle down financially, you were wrong. That means you, Hewlett-Packard. And especially you, IBM . In the quarter ended November 30, Oracle's overall revenues were up 47 per cent, to $8.58bn, and not just because of the addition of Sun's hardware …
Oracle is having short term success because they:
1) raised hardware prices
2) raised maintenance costs and forced everyone to put service on every box
3) took sales direct
Current:
1) customers are getting off Sparc as fast as they can
2) Oracle is being sued for its monopoly position and abuse
3) customer satisfaction is the worst in the industry
Future:
1) Oracle has dumped Fujitsu so the highend market is cratoring and they cannot sell enough Exadatas as they are $10M per rack. There is a reason they don't talk boxes....they still have not sold 100 yet. They talk $2B pipeline...well Larry sooner or later your exaggeration will come back to you because you can't keep lying. As for Mark Turd...I like how you make up your numbers on the fly 60/40....nice answer.
2) Customers leaving will quickly erode the maintenance stream
3) Customers will go to Power and Nehalem first then DB2, SQLserver and Postgres next
In the end:
Same fate as Wang, Digital, Montgomery ward, Circuit city, etc..
You piss off your customers and they never forgive you.
Unfortunately, your Sunshiner blinkers meant you misread a leap in Snoreacle software profits as meaning SNoreacle was actually selling more of the old Sun kit. The reality, as reported here http://www.channelregister.co.uk/2010/12/02/idc_q3_2010_server_numbers/ show that Pee7 and Tukzilla are doing just fine, but Snoreacle is struggling to grow numbers despite the economic upturn. The article also concludes that the real drivers in servers are Windows and Linux, definately not Slowaris, and fields in which Snoreacle lags even Dell, let alone IBM or hp. The inescapable conclusiuons are that customers still don't want the old Sun servers even with added Larry Bluster (TM); that the old Sun installed base is shrinking and being replaced by non-Snoreacle kit from other vendors, which means the Snoreacle support biz is also going to shrink; Snoreacle is being outgrown in server shipments by whitebox suppliers, and so realistically has zero real chance catching IBM or hp.
Isn't funny how the Sunshiners have all come out of hiding now they think there's some good news about Snoreacle?
/SP&L
"Oracle is not allowing HP to publish benchmark results based on its eponymous database running atop HP-UX, HP's variant of Unix. Why HP has not put out results for a full-bore Superdome 2 machine running Microsoft's Windows and SQL Server combo might seem like a bit of a mystery, but Microsoft is backing off from supporting Itanium with future Windows stacks and HP doesn't want to bring that up."
Oracle strangling HP, even MS pulling itanium support! RHEL 6 with no itanium support, not looking good for the Superdome's is it.....
When you have the vendors of the worlds dominant operating systems pulling support for an architecture, it's done.....end of.
"....When you have the vendors of the worlds dominant operating systems pulling support for an architecture, it's done...." Please post the Niagara, UltraSPARC, or Snoreacle SPARC64 servers that come with Snoreacle support for RedHat or Windows of any form. And you can still run current versions of Windows Server and RHEL on hp's Integrity Itanium servers with full support from hp and M$ and RH, it's just you won't get the new versions.
BTW, did someone forget to tell Larry that over 50% of all new Oracle instances go on hp servers? Maybe that's because hp can supply a more complete offering rather than just Larry's tiny and walled garden with it zombie server and storage products.
/SP&L
For how long? Does not sound like a good investment, if you ask me .... :-P And when will Intel stop throwing more money at expensive platform development nobody is using, anymore ... exept HP?
HP-UX = [Defunct] OS
There, done!
PS: Matt, you had not yet replied when I was writing my last comment ...
Oh dear, it's alwasy amusing to see the tragic delusions of the Sunshiner laid so bare!
"For how long?...." Minimum ten years, even if they announced a halt to hp-ux today. And hp assure me they're getting ready for the next hp-ux 11i release next year, so more than ten years after that. Meanwhile, Snoreacle are dropping releases of Slowaris, locking down MySQL and and screwing over the Java community.....
"....And when will Intel stop throwing more money at expensive platform development...." But Intel makes massive profits and has co-development between Xeon and Itanium, which means Itanium shares from the massive advantage of the economies of scale of Xeon being the most popular server CPU. On the other hand, Niagara/CMT shares nothing with anything, has tiny share, and Snoreacle can't afford the development (because they'd have to replace half the development team that Sun fired as cost cuts for a start). SPARC64? LOL! Fudgeitso hasn't commited to anything beyond the current generation, depsite Larry's announcements, and Larry hasn't the ability to replace it seeing as all he inherited from the Sun carcass was the left-overs from a CPU team that failed with Rock. Looks to me that Itanium is a much safer development than anything coming out of Snoreacle!
"....HP-UX = [Defunct] OS....." <Sigh> Maybe you should try talking to someone that actually works in enterprise IT? The reason Red Hat decided to drop future developments on Itanium was because they weren't getting a big enough return as customers were choosing hp-ux over RHEL. Same goes for Windows on Itanium. Both RH and M$ decided to concentrate instead on x64. The IDC figures show Tukzilla is ramping up faster than Snoreacle's whole hardware biz, and if the number going out with RHEL or Windows is dropping that means more customers are buying more Integrity with hp-ux. Now, would you like me to repeat that in smaller words or maybe draw you a picture?
/SP&L
I got this, yesterday, in a sourceforge email, sponsored by Novell ...
http://p.sf.net/sfu/novell-sfnews
Novell is basically telling Solaris users to switch to ... Suse, mainly because Oracle turned its back on the community ... what they seem to forget is their Microsoft deal, which meant Novell turned its back on the community, too ...
Matt Bryant's wet dream ...
I do not like Oracle much, but as a Solaris fanboy, I'd love to see Oracle crush HP and their [defunct] OS.
I said it before, and I will say it again.
"Sun had 35.000 paying customers. Oracle has 350.000 customers. If Oracle makes a tiny percentage to switch, then Solaris is gaining whereas rest of Unixes is taking a beating by Linux. Earlier there were only technical advantages to switch to Solaris, but Larry will make it sure that there are also business and economical advantages. The Solaris future is bright"
Well see in a few years how this trend will continue. Oracle DB runs on Solaris, more than any other OS. This trend will be fortified.
Solaris rulez. Why? Because it is best. ZFS, DTrace, etc. :o)
"....With Linux, HPUX, AIX, Windows you must purchase a separate filesystem for RAC as well, if you want a FS...." Nice try, Bill, but not quite 100% true. For starters, hp-ux comes with LVM bundled in, not a separate purchase, and that's fine for RAC. You can get such extras as the Veritas Cluster File System if you prefer, but that's usually becasue you want compatibility when migrating off all those old Slowaris boxes (which HAD to have Veritas to work in clusters properly and reliably) onto Integrity. And ASM? Not if I can help it!
As regards Linux, I'd rather use Red Hat's GFS, or Polyserve (can't remember what hp are calling it nowadays), or (gun to my head to admit it) just NFS of NetApp which seems to work quite nicely. "Free" options include ext3 (fully-supported by Oracle), ext4 (not sure if this is supported yet, haven't checked in a while), or ReiserFS (SLES only, IIRC). I'm told BTRFS also works quite well but Oracle won't support it as they want to sell you an OCFS licence.
ZFS will be supported. They are most probably working on it. It takes some development time to do that. Oracle bought Sun quite recently.
But, that does not take away the superiority of ZFS. ZFS is still awesome and protects your data, which no other filesystem does:
http://en.wikipedia.org/wiki/Zfs#Data_Integrity
How exactly does ZFS "protects your data" in a manner in "which no other filesystem does"? Are we talking software RAID, something that predates ZFS by years and was even possible with a hack of Windows XP long before ZFS arrived? Just what is this magical, ZFS-only capability? Hate to bring some reality into your life, but ZFS is just a WAFL clone, so anything ZFS can do has already been done by NetApp and probably done by a dozen other conmmercial and free products long before ZFS.
End to end data integrity. Look it up Matt. No other general purpose filesystem does this (even WAFL). To be fair WAFL does do a basic WAFL, but not to the same extent that ZFS does. You may write the data to disk, and it may be mirrored, but did you know that it was written wrong, and that data was written wrong on both mirrors? ZFS knows it. WAFL sometimes knows it. No other filesystem knows it.
Here's just one of the studies done on the subject Matt:
http://www.cs.wisc.edu/wind/Publications/zfs-corruption-fast10.pdf
"End to end data integrity...." Welcome to the new Snoreacle-Sunshine feature sell! Yes, yet again it's a non-feature that Snoreacle will hype and try and convince customers they MUST HAVE, that all comparissons must be made on the basis of "does it have end to end data integrity?" What a complete load of cobblers. You can already see this rubbish being sprouted all over the Snoreacle marketeering sites and repeated in the Sunshine blogs. Jeff Bonwick's is a perfect example, starting with the following simpleton statement:
"The job of any filesystem boils down to this: when asked to read a block, it should return the same data that was previously written to that block...."
Yeah, Jeff that's why we have RAID. Hardware raid in the case of internal disks, which means no lost CPU cycles as a bonus. He carries on with more twaddle:
"....Incredibly, most filesystems fail this test. They depend on the underlying hardware to detect and report errors...." That should be they don't NEED this feature BECAUSE they depend on the underlying hardware. Well, they depend on reliable hardware and good designs in other vendors like hp, Dell or IBM, but not so much of a surprise that ex-Sun customers might think they need a bit more. In short, Snoreacle is trying to pretend there is some great hradware problem that only ZFS can solve. Anyone with half a clue can see through this junk. Anyone with any experience of real enterprise hardware design would simply laugh Jeff out of the room. I really hope for any of you tasked with flogging the new Snoreacle kit that this is not your opener, otherwise you're going to be shown the door almost as quickly as if you were selling Philip R Greaves's "Pedophile's Guide to Love and Pleasure".
"....That should be they don't NEED this feature BECAUSE they depend on the underlying hardware. Well, they depend on reliable hardware and good designs in other vendors like hp, Dell or IBM, but not so much of a surprise that ex-Sun customers might think they need a bit more...."
Look Matt. You can not trust the underlying hardware. Have you seen the spec of a new enterprise SAS disk? No? Well, it typically says "1 irrecoverable bit in every 10^16". Just check for yourself. For instance, this Atlas SAS Enterprise disk:
http://www.seagate.com/staticfiles/maxtor/en_us/documentation/data_sheets/atlas_10k_sas_datasheet.pdf
"Data Error Rate per Bits Read
Recoverable <10 per 10^12
Nonrecoverable <1 per 10^15" <------- OBS!!!!!!
.
.
Matt, I think you should study this fact of "Silent Corruption" more. CERN has done several studies on this. You know, CERN stores lot of data from their Large Hadron Collider. The experimental data costs billions and decades to extract. What happens if some bits are flipped silently? Have you heard about bit rot? See here what CERN says, about their unsafe hardware raid solutions. Here are links to CERN studies, you can read the CERN studies yourself:
http://storagemojo.com/2007/09/19/cerns-data-corruption-research/
.
.
Matt, I suggest you google on "Silent Corruption". You are in for a surprise. Here you see five different research papers that show hardware raid might corrupt your data:
http://en.wikipedia.org/wiki/Zfs#Data_Integrity
And also, ordinary filesystems such as XFS, JFS, ReiserFS and NTFS might corrupt your data. There is a PhD thesis that shows this. Just read this link. Or do you distrust all researchers? In that link there are several research papers that show hardware raid is not safe. You can not rely on your hardware raid to protect your data. Your data is unsafe, Matt. It is not Oracle propaganda. Independent computer scientist researchers say this. Just read the above link and you can find all the research there.
Ah, poor Kebbie, Snoreacle are obviously using him as a lab rat to try out their marketeering material on!
"....You can not trust the underlying hardware..." Erm, actually, yes I can, because we have redundant hardware, and keep redundant copies of data. Where there is particularly sensitive data we have three live copies and can run compares to spot the dodgy one, and logs to go back and rebuild if necessary.
"....Have you seen the spec of a new enterprise SAS disk?..." Yes, I've seen the MTBF numbers, run thrash'n'burn tests on arrays, and looked at the data. A SAS disk is a far lot less likely to fail than a software solution. In fact, the most common cause of corruption is an operator action, such as when an Oracle DBA drops the wrong table by accident. Your "end to end data integrity" would be just useless against that DBA's mistake.
"...I think you should study this fact of "Silent Corruption" more...." Been there, got the T-shirt, still not seeing the need. You see, we have solutions that work, whereas you are pushing ZFS which - frankly - has a shocking record of falling over and corrupting customer data. The Web has lots of stories about ZFS crashes and corruption. You can claim it's all fixed now but I think I'll let someone else take the risk seeing as my own solutions are known to work.
"...ordinary filesystems such as XFS, JFS, ReiserFS and NTFS might corrupt your data..." Euugh, is that the smell of troll manure? Please supply some data on how ZFS has managed to make iteslf so much less likely to corrupt than Resier or ext, especially as both the latter have solid track records of reliability compared to the bugfest that is ZFS. And when you get down to it, RAID-Z is just software RAID using unused diskspace - not revolutionary, been done by WAFL for years.
And here's the big problem with ZFS - it needs control of disks and filesystem, which means one problem with the OS and you have a mess on both. One patch means everything has to come down, and then one bug means everything will go down. With hardware RAID, which is basically software RAID happening on dedicated and separated components (a "mini-server" if you like as that's what array controllers are), if the OS screws up then the disks are segregated and can be presented to other servers. One LVM import or disk scan and you're back in business. But as ZFS is a local filesystem only and can't be clustered it is a SPOF. Once that local OS has screwed up it means any other servers accessing the data (unshared data, that is, becasue ZFS can't do sharing) lose access too, which means it is a very unresilient design and not suitable for highly-available or biz critical solutions. Which is why it's not supported with RAC. So even if your "end to end data integrity" was even vageuly necessary, I'd be looking for a different solution as ZFS is just not mature and doesn't have the features needed for enterprise solutions.
""....Have you seen the spec of a new enterprise SAS disk?..." Yes, I've seen the MTBF numbers..... A SAS disk is a far lot less likely to fail than a software solution."
Matt, you dont seem to understand what I try to explain. You talk about MTBF on disks, and they will not crash in another 10.000 hours or so. I do not talk about disk crash.
I talk about Silent Corruption. Have you heard about ECC RAM? It protects against bits flipping randomly in your RAM memory sticks, caused by cosmic radiation, etc. ECC does not protect against Memory sticks crash and break. Bits in RAM flips on random, you did know this, yes? What do you think is the purpose of ECC RAM? I hope you dont mistake the probability of RAM memory sticks crash (MTBF) with the probability that bits flip on random (ECC).
Have you heard about Bit Rot? It causes bits on disk, to flip randomly. Bits start to decay after a while. If you need ECC RAM, why dont you need the same on disks?
Yes, you have checksums on disks, but they are not good enough to correct all bits that spontanously get flipped on random on disk. As I showed on the spec on a SAS Enterprise Atlas disk, one bit on every 10^15 does not get corrected. This is from the spec sheet from the manufacturer. It is NOT about the disks crash MTBF. The disk continue to work, but bits have spontaneously flipped on random. You talk about MTBF, I talk about Silent Corruption.
And you know Physics Centre CERN talks about silent corruption. I hope you consider CERN have some credibility? CERN says silent corruption is really problematic.
http://www.zdnet.com/blog/storage/data-corruption-is-worse-than-you-know/191?tag=rbxccnbzd1
.
.
"..."...ordinary filesystems such as XFS, JFS, ReiserFS and NTFS might corrupt your data..." Euugh, is that the smell of troll manure? Please supply some data on how ZFS has managed to make iteslf so much less likely to corrupt than Resier or ext,..."
We have supplied you the data you ask of. I have posted links and research papers on this. I can post them again, and this time I will quote text.
Here is a PhD thesis which I quote:
http://www.zdnet.com/blog/storage/how-microsoft-puts-your-data-at-risk/169
"Dr. Prabhakaran found that ALL the file systems [NTFS, ext3, JFS, ReiserFS, XFS, etc] shared
. . . ad hoc failure handling and a great deal of illogical inconsistency in failure policy . . . such inconsistency leads to substantially different detection and recovery strategies under similar fault scenarios, resulting in unpredictable and often undesirable fault-handling strategies.
. . .
We observe little tolerance to transient failures; . . . . none of the file systems can recover from partial disk failures, due to a lack of in-disk redundancy. "
And you also asked for proof that ZFS protects your data better. We have posted that research paper here earlier. Here you have a link, again. You can find the research paper on this link. And I quote from the research article by the computer scientists:
http://www.zdnet.com/blog/storage/zfs-data-integrity-tested/811?tag=rbxccnbzd1
". . . ZFS successfully detects all corruptions and recovers from them as long as one correct copy exists. The in-memory caching and periodic flushing of metadata on transaction commits help ZFS recover from serious disk corruptions affecting all copies of metadata. "
.
.
"...And when you get down to it, RAID-Z is just software RAID using unused diskspace - not revolutionary, been done by WAFL for years...."
No. RAID-Z is not just ordinary raid. It is something unique and different. It protects your data. There is no chance in hell that WAFL can do something like this:
http://jforonda.blogspot.com/2007/01/faulty-fc-port-meets-zfs.html
.
.
"...And here's the big problem with ZFS - it needs control of disks and filesystem, which means one problem with the OS and you have a mess on both...."
Actually, that is THE advantage of ZFS. This means that ZFS is in charge of the whole chain. From RAM down to disk. End-to-end in the chain. That is the reason ZFS can detect errors that WAFL can not (read the link).
BTW, if you use hw-raid card together with ZFS, then you have a problem. ZFS can not guarantee data integrity if you use hw-raid. ZFS gets confused and your data might corrupt. hw-raid is a no-no with ZFS.
.
.
I have basically posted, and quoted this text. Before we continue, can you just read this text? Or do you prefer me to copy& paste here?
http://en.wikipedia.org/wiki/Zfs#Data_Integrity
Kebbie, before we debunk the rest of your junk, please try and THINK about the realities of the workplace before posting such twaddle.
".....Hw-raid does not protect data...." It does. Simple as. Which is why hardware RAID is used and depended on by millions of businesses Worldwide. Please deny that your own business is not using it right now.
Bit rot, silent corruption, they're not new, they just something the Snoreacle marketeers have seized on in an attempt to flog ZFS. The bigger problem for ZFS is, as you have already admitted, it can't be clustered and can't work with hardware RAID. I can live with the small chances of silent corruption and bit rot, I can wortk round them, but I can't implement the systems I need with ZFS as it is a great, big SPOF (and a bugfest to boot). If I introduced ZFS into our solutions I would be massively INCREASING the risk of data being corrupted or production being lost. In short, carry on plugging your feature-sell, just don't bother round here 'cos we already see through it.
/SP&L
Well, as AC stated above, it's not like Oracle haven't put a lot of ships in the sea to make more money of it's SPARC product line, and it seems that it's paying off. BUT, according to the IDC numbers I've seen then the Oracle UNIX revenue have only risen 1.4% from Q3 2009 to Q3 2010.
And 1.4% is not in the same league as 46% on the software side, so sure it's good for Oracle, and for UNIX customers, that it's still a kind of 3 horse race.
BUT it's not a momentum change IMHO.
// jesper
For someone hyping themselves as understanding Enterprise platforms you should really know better Matt. The platform is there to run the app, it's not the platform driving the app choice and a very large part of the choice is the App licencing cost. You & I both know Oracle licence costs are extortionate & make the platform cost a secondary consideration. If the platform was the primary choice why the hell would people choose Linux over any of the mature Unix's like Solaris, HPUX or AIX? Yet more and more people do.... (Let's not divert into a Linux debate if we can help it Matty boy).
Unless you've got in-house code locking you to an OS / architecture then App licencing costs drive the platform choice so you've got to accept the reality that Oracle genuinley have whacked HP with by the 1.0 multiplier, it makes Intanium really expensive for licencing.
Please don't bleat about more cores, Itanium got some more cores, so bloody what! Did you notice the 8 -> 16 core count increase on the Sparc CPU's, did you fail to notice the leaps in core counts that AMD & Intel X64 are doing? Did you notice the cores and the frequency of the cores in the IBM series and the perf stats ?There's talk of Intel / AMD's with _24_ cores per socket coming through so Itanium looks even weaker. 24 cores Matt! It's simply cheaper to licence Oracle on anything except Itanic.
Anyways Matt, I guess you frothing now and possibly 5'000 words through your rant & I'm the devil but it much more sense (financially) to run Oracle on Sparc & very attractive to run Oracle on Intel / AMD @ 0.5.
Yes, HP sells X64 gear, so does Oracle, IBM, Dell, CISCO, yadda yadda blah blah....
Still pointing and dreaming Matt?
PS: In our Bank all the HP DL's were dropped and it's all IBM x series. HP don't have it sown up.
"For someone hyping themselves as understanding Enterprise platforms..." Not just "understanding", that's "working with on a daily basis", which is probably something you don't do.
".....The platform is there to run the app...." Correct! Shame the rest is complete twaddle. The platform has to be a stable, supportable and predictable platform. When chosing a platform you look at the prior history of the vendor and the relationship you have had with that company as well as just performance figures. The problem for Snoreacle is most customers see "Oracle = software", and "Sun = hardware failure". Sun burned through just about all the goodwill generated by the good old days when SPARC-Slowaris was actually respected and trusted. Customers looking for a platform today to run biz crit apps on will consider IBM and hp much better choices, simple as that.
"...If the platform was the primary choice why the hell would people choose Linux over any of the mature Unix's like Solaris, HPUX or AIX?....." The problem for that complete pile of fail is that market figures show that Linux is not replacing UNIX wholesale. It is replacing a lot of Slowaris as ex-Sun shops simply don't have faith in Snoreacle's ability to deliver. But AIX and hp-ux seem to be very resilient to Linux. The growth area for Linux is still on x64 in replacement of or as alternative to Windows, not UNIX.
"....Unless you've got in-house code locking you to an OS / architecture then App licencing costs drive the platform choice...." Erm, no. Actually, what drives the enterprise platform choice is risk reduction, and Snoreacle is seen as risky. Customers that have enjoyed good relationships with hp and IBM (as we have), and seen a good relationship with Sun turn sour (as we did), will pay the extra to buy hp or IBM. This is very simply proven by the fact that Sun was selling solutions CHEAPER than IBM or hp during the final years of the Sunset, but still sales crashed because the customers bought the more expensive hp and IBM kit. Please try and explain how your theory can stand when the evidence of Sun's demise shows it is rubbish?
"....Please don't bleat about more cores...." Why, because it completely explodes the twaddle you present? Price per CPU doubled, but core core count also doubled, so the effective change is ZERO. Well, actually, seeing as the new cores are more powerful, it means the same licence costs for more power. You can try and cut it any way you like but you're still denying what is patently obvious.
"....24 cores Matt!...." Yes, but what will those cores be capable of? Xeon already has eight cores, but we buy Itanium kit because our PoCs show it outperforms the RHEL/Win/Xeon option, because the individual cores are more powerful. This also explains why Niagara is such a failure - large core-count but cores that can't compete. Please try and pretend it isn't so, but at the same time please explain why Niagara sales are a pitance compared to Xeon if mutli-core is the answer to everything? Oh, that would be because the individual power of each core is still relevant, and Tukzilla still has massive per-core advantages over Xeon or Opteron. Sure, there are cases where a 24-core Opteron will be the better option, just as there are a few cases (like webserving) where Niagara makes sense, but the market has shown that the majority of enterprise apps require real cores, AKA Pee7 and Tukzilla.
"....but it much more sense (financially) to run Oracle on Sparc...." Financially, as long as you don't consider (a) you will need more SPARC cores, and (b) you massively increase the risk, as shown by the history of Sun. And then, when those points are considered, you'll find the board buys hp or IBM. Or even Dell if it's x64, rather than Snoreacle.
"....& very attractive to run Oracle on Intel / AMD @ 0.5...." Which is surely an indication of why Snoreacle will lose the hardware battle? After all, Snoreacle is third tier when it comes to x64 and being out-sold by the whitebox vendors! Any sales of Oracle on x64 will more likley be on hp hardware, and that is simply a fact.
"....Yes, HP sells X64 gear, so does Oracle, IBM, Dell, CISCO, yadda yadda blah blah...." You forgot that hp sells a lot more x64 than anyone else. A LOT more. And that all has services and storage and software sales pulling theough with it. Not to mention the printers, which I know really uspets you frothing Sunshiners.
"....Still pointing and dreaming Matt?...." No, still pointing and laughing at you Sunshiners.
"....PS: In our Bank all the HP DL's were dropped and it's all IBM x series. HP don't have it sown up." So, not Snoreacle x64 boxes then? There's a surprise - not! I'm also amused to see your example of an "enterprise" swap-out is purely Wintel. Guess that exactly places the limits of your enterprise experience.
/SP&L
"....this also explains why Niagara is such a failure - large core-count but cores that can't compete. Please try and pretend it isn't so, but at the same time please explain why Niagara sales are a pitance compared to Xeon if mutli-core is the answer to everything?..."
Maybe Niagara sales are lower than Xeon, because of the same reason Itanium sales are lower? More people use x86 than Niagara or Itanium.
And when you say that Niagara is slow, well, it holds several world records. I have posted links that show world records for you earlier, but can repost them for you again if you wish.
Also, recently, the TPC-C world record of 30 million tmpc was done on Niagara running "Slow-laris".
If I am not mistaken, HP's best TPC-C record is 4 million tmpc? So how can you say "Slowlaris" or say Niagara to be "slow" when it is quite, much faster than HP's fastest offer? What you say does not make logically sense? Niagara running Slow-laris is almost 10x faster than HP's best offer in TPC-C? You call that slow? What do you call HP's best result of 4 million tmpc then? "HP is ultra-slow"? How much faster must Niagara+Solaris combo be, for you to say it is fast? 100x faster?
<Yawn> Kebbie, you can sprout cherrypicked benchmarks from Sunshine lab sessions with zero real-World relevance until you're blue in the face, it won't make any difference. The market has spoken and that's what killed Sun. Simply trying to rehash the same gumph with a heavy does of Larry's Bluster (TM) is simply not fooling anyone. Niagara does have its sweet areas, but the problem is those areas (like webserving) are also prone to being good for cheaper x64. Niagara has zero penetration into the high-end as it was never meant to play there, Rock was. Niagara was supposed to stop x64 eating the Sun base from below, so comparisons with x64 sales are quite valid. Please stop trying to pretend Niagara is even vaguely competitive with Power or Itanium with real World applications as we all know it's simply not true.
"Maybe Niagara sales are lower than Xeon, because of the same reason Itanium sales are lower? "
What HP have done is to drastic reduce the price of SuperDome's with a factor of 4-5 for the same performance. I tell you that is something that can be seen on sales numbers.
IBM have done much the same, as they are selling POWER7 systems that are 4-5 times faster than their POWER6 counterparts at the same price or less.
Oracle are basically selling servers at the same price, so for the M series it's ... well.. 20% more bang for the same buck. IMHO M series just isn't competitive any more with HP and IBM offerings.
T series is basically sold at the same price as before, so it's a factor of 2 better in price performance.
"Also, recently, the TPC-C world record of 30 million tmpc was done on Niagara running "Slow-laris"."
It's a cluster... CLUSTER. says nothing about solaris scalability or performance. Yes each server in the cluster does around 1M tpmc for a 64 core server.. That is about 25% of the nodes that IBM or HP could have used.. 3 1/2 years ago. The only real good story about the submission is that the software could scale to 27 nodes, which I find really terrific. But lets substitute those 27 nodes with POWER 780's or SuperDome2's.. and then you'll see throughput. Wroom Wroom.
Now if you could administrate 27 T3-4 as a single system image, then you'd have to use the same tools as used on BlueGene or SGI's Altix'es. Cause there would be 13.824 hw threads to keep track of. *BOGGLE*
// jesper
Ok, so HP and IBM has reduced their prices, whereas Oracle has not?
This only proves one thing: competition is good. Without fierce competition, IBM and HP would not have cut their prices. So this helps you, as an IBM customer. Without competition, IBM would have their Unix prices in the stratsphere, just like IBM Mainframes. IBM Mainframe marginals are extreme, much higher than Unix servers.
Also, you have forgotten that Sun never was expensive. Sometime IBM was like 10x as expensive or more, for the same performance, that you got from Sun. I am thinking of the SIEBEL v8 benchmarks, where you need six POWER servers to match one Sun T5440. If I am not mistaken, one P570 server costed 413.000 USD. And the single T5440 costed 76.000 USD.
Or, for instance, this benchmark
http://www.zdnet.com/blog/murphy/p570-vs-t5220/1537
Here Sun only had a six times price/performance advantage, though. This presumes you got one IBM server for free.
So, yes, I think it is good that Oracle exists and IBM and HP is shit load of afraid of Oracle. Otherwise IBM would still have sky high prices. Come on, 413.000 USD for a single slow server?
Thanks to Oracle, IBM and HP has cut their prices. I understand Oracle has not cut their prices, they are already cheap.
.
.
Regarding the TPC-C record that Oracle have. Yes, it is a clustered result. So what? They are fastest in the world, by far. If you need the most extreme TPC-C result you have no other choice than Oracle. I dont think Wall Street Banks and Hedge funds care if a cluster is delivering the extreme result?
.
.
"....Yes each [Niagara T3] server in the cluster does around 1M tpmc for a 64 core server.. That is about 25% of the nodes that IBM or HP could have used.. 3 1/2 years ago..."
25%? So IBM and HP should be 4x faster?
And you also talk about POWER780 or SuperDome2 running a cluster to get even higher record than 30 million tmpc. Good luck on that. If IBM or HP can get their gear to scale just as easy as you say, then surely they will soon break Oracles record. So, IBM should present a result that is 4x times higher, that is, 120 million tmpc soon.
I dont think IBM will in 2011 present a result of 120 million tmpc. But you seem to be convinced IBM will? "Simple matter of inserting IBM servers instead of Oracle servers" - or something like that? I dont think it is that simple. To scale good is not trivial.
But we will see. If IBM presents 120 million tmpc, I will admit you were right and I was wrong. But I am convinced that will not happen.
You seem to forget that it's HP and IBM that are cutting their prices not Oracle. And it's SUN that traditional have had the biggest UNIX marked share.
And I don't think your comparison of obsolete quarter/half filled POWER 8 socket servers versus fully filled obsolete Oracle servers with 2/4 sockets, is particular relevant.
Try to compare T3 versus POWER7, these are the servers that people are buying, today. People are not buying POWER6, why should they, when they can get x5 the price performance by buying a POWER7 machine.
And you still don't get the difference between a cluster and a single SMP machine.
// Jesper
I am not forgetting anything. I am saying you that IBM and HP are cutting their prices to be more competitive with Oracle. With Sun, HP and IBM was not afraid. They did not cut their prices. But Oracle is a major threat. HP and IBM must cut prices. This is a testament to how scared they are of Oracle. And you, Jesper, benefit from their fear. You can buy cheaper hardware. :o)
.
And I still dont get the difference of a cluster and a single SMP server? What has that to do with Oracle having the TPC-C world record? Yes, I know it is clustered, so what? If you need the highest TPC-C performance (Wall Street funds and banks), you have no choice but Oracle. They dont care if it is a cluster or a single SMP server - do you think they care?
Whoa there Matty, you bit and swam so fast I almost ran out of line, lets start the play to get you in shall we? This ones coming in fighting!!
Cutting through the first two paragraphs of self grandeur followed by the usual anti-Sun bashing (true HP fanboy rants, did you notice Sun doesn't exist?) we finally arrive at some meat! Matt, I'm not pro-Sun, I think Redhat have stolen the Market at every other Vendors expense and will keep on doing so. Linux is just going to improve & keep eating at expensive RISC Unix boxes from all vendors but lets not deviate. I find it immature compared to most Unixes, just look at it's VM management, poor until RHEL6, it's schoolboy but anyway. I guess I have no experience as you say!
You may have noticed the trend that when a vendor looks like it might go bust people stop buying. It's not like they had a sudden downturn in sales, it slow and protracted but I think Sun were still on top of HP in terms of Server shipments and still held out over HPUX in terms of Solaris. You could be honorable and admit a similar nervousness exists around an architecture close to your heart as people feel it's on it's death bed but I doubt you will.
You do seem to tune into this sentiment to your advantage by the next paragraph so I'll let the comment stand. When people are nervous or uncertain they won't buy, simple as that.
Then you bleated, I tried Matt, I really tried to steer you away from stupidity! The point your trying hard to ignore is while you feel it's OK to maintain the same bang-per-buck on Itanium (cores doubled, so did the licence cost) everyone else is getting more value. AMD / Intel core count is increasing so the licences get more bang per buck, Sparc actually got cheaper and the roadmap is doubling the core count so that by my book makes two rounds of bang-per-buck discounting on Sparc.
I suppose it works in your mind, if you just close your eyes and repeat something long enough maybe you'll believe it. I think a lot of cults work that way, anyways, onwards.....
With respect to workload carried out by CPU's we all know Xeon / AMD are very capable of carrying a respectable workload & so is Itanium, Power and so forth. Sparc is getting some serious investment even though you downplay it Matt but it's not about Sparc. You need a slap as to whats happening to Itanium.
The ISV's left a while back Matt, no more RHEL, no more Windows? You think the HPUX and nonstop sales are enough to cover the R&D costs? Do you really believe enough components are shared to allow it to survive? I give it 18 months till a HPUX port to x64 occurs. Anyways, about on cue now we have more frothing.....
In terms of financial sense, don't you think Oracle will pitch entire suites of Server, OS and App fee's at people instead of allowing people to walk off and do a hardware purchase elsewhere. They won't stop people but there going to make it attractive. It's not illegal, IBM will discount for the sale and discount to get DB2 into places. Whose got financial sense really Matt?
With respect to HP additional services Matt you seem delusional. Oracle have a massive suite of software, so do IBM. HP is a server seller with some management tools of dubious quality, I'll let rest there. When they offer an Enterprise DB, App servers and software for specific business purposes come back. Don't try to compete Matt, you know it's going to be embarrassing to look at HP's software offerings but hey ho, go for it. Could be an amusing comparison.
And the final paragraph? Matt, thats trolling but I guess I fed you. I'm not saying who I work for, it wouldn't be appropriate but it's not a small place, lets leave it there. Windows is a small part of the install base and is used for managing the desktop needs (sharepoint, mail, some SQL) but everything serious is on Linux, some Unix or Mainframes and guess what? HPUX is a legacy OS used in one arm of the business & is being replaced with Redhat. On IBM. No more bank build for HPUX I'm afraid but feel happy knowing Engineering cut a Solaris build recently as well as the regular churn out of RHEL builds.
Sorry Matt, you were hoping but your wrong.
Keep pointing and dreaming, the doctor has some drugs for that I think!
Your fishing arm seems a bit weak so I'll keep it short and give you time to get your breath back.
"....I'm not pro-Sun...." Yes, and all those indicators of the typical Sunshiner, they're just coincidences, right?
"....I think Redhat have stolen the Market at every other Vendors expense...." Exhibit A - hp and IBM partnered with the Linux community and made money from Red Hat, especially when it came to attacking the massive SPARC-Slowaris base. Sun went head-on against Linux and died. The fact you consider Red Hat to have "stolen" anything shows the anti-Linux mindset of a typical Sunhsiner.
"....I find it immature compared to most Unixes, just look at it's VM management...." Exhibit B - having pretended at a liking for Linux, you then sprout the typical anti-Linux FUD, usually to be found in diatribes around the "superiority" of Slowaris Containers.
"....You may have noticed the trend that when a vendor looks like it might go bust people stop buying...." Typical Sunshienr evasion of reality - Sun started to look like it might go bust BECAUSE customers stopped buying over several years. Depsite what you Sunshiners like to pretend, Sun didn't just go bust by some strange happening, it happened due to a period of rapidly declining sales due to failed developments and poor support leading to lack of customer faith. Us customers stopped buying the Sun kit and bought other vendors instead - period!
"....but I think Sun were still on top of HP in terms of Server shipments and still held out over HPUX in terms of Solaris..." Exhibit C - the complete detachment form reality typical of the Sunshiner. Sorry, but the IDC computer says "No!" Same for the Gartner one. And going on about the number of Slowaris installs out there - mainly old UltraSPARC and therefore very vulnerable to Linux - is just classic Sunshiner.
"....Sparc actually got cheaper and the roadmap is doubling the core count..." Exhibit D - not a Sunshiner but leaping to defend SPARC? Why not Pee7, I wonder - not.
"....You need a slap as to whats happening to Itanium...." Well, the Reg posted an article on IDC's Q3 figrues which said Tukzilla is ramping up just fine, whilst Snoreacle server sales are being outgrown by whitebox vendors. I suggest you go slap yourself.
The rest of your laughable post is juts the usual Sunshiner FUD about how long can hp maintain Itanium with only hp-ux and NonStop (a lot longer than Sun could Slowaris, by a year already!), with the usual factless assumptions, and then the new Sunshienr line - Snoreacle can supply the whole stack, IBM can too, but can hp? Of course, that completely avoids the fact that hp can supply several stacks using partners like SAP, IBM, RH and M$, which is real choice compared to Larry's walled garden. It also completely ignores the fact that Larry sells more new Oracle licences as part of hp server sales than any other vendor's, and more licences into the enterprise market on hp-ux servers than any other platform. If you disagree with that then I suggest you take it up with Oracle, as they had to admit that to me.
Yeah, you're so obviously not a Sunshiner. Oooh, look - a flying pig! Get back under your new Snoreacle bridge, troll.
/SP&L
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
