I believe, I believe
I believe everything Larry says.
Larry Ellison promised to hang onto Sun's hardware business yesterday, saying he wanted Oracle to be like Apple and Cisco. The firm filed a canned interview with Ellison with the SEC yesterday, in which he said "we're definitely not going to exit the hardware business." Ellison told Reuters via email that while most hardware …
I believe everything Larry says.
Oracle's difficulty with SPARC is this.
1) If they keep quiet, the competition will spread FUD that Rock will be canned, SPARC is a dead end.
2) If they say too much about Rock, customers will defer Sparc64 purchases wanting the newer stuff.
Sun's approach in the past has been to give little information about upcoming systems, even under NDA, until close to launch date in order to prevent (2). This statement from LE is meant to counteract (1) without the risks of (2).
There are some very, very strong technical synergies between Oracle and SPARC/Solaris. The most difficult part will be Oracle moving from one state of certainty (RAC on GNU/Linux is the perfect solution) to a different state of certainty (Single instance on CMT SPARC is the perfect solution) without being seen changing their mind.
"Oracle would stick with the SPARC architecture, he said, pointing out that "even Apple is designing its own chips these days".
Except where they aren't and, er, never did, well not CPUs anyway... In fact unless I'm much mistaken Sun used to use Motorolla 68000 series chips back in the day (albeit the very sweetest batches only) and, like apple, eventually ended up selling x86 boxes partly because it's widely regarded as NUTS to try and make your own processors these days.
I'm sure Oracle love the thought of the lock in that would get them though!
I can't recall a time...
I'm thinking of starting up a betting pool. How long will it take for Oracle to do an about face and dump most of the hardware business?
There is no way that SPARC innovation, based on limited production volumes can keep up with Intel/AMD R&D; they are being bypassed at a rapid pace by the competition.
Then, there is the issue of margins. How high does he think that he can raise the margins of hardware; can it ever approach the margins of software? I don't ever see that happening.
Wait until Oracle gets the wakeup call of HW competitors such as IBM, HP, and Dell making a major push to recommend other software than what Oracle is selling to their customers. Why would they want to subsidize a competitor? Customers of non-Sun hardware may also be wary of purchasing software from a different hardware vendor for their systems.
The smart thing is to dump the HW, sell SPARC for whatever you can get to Fujitsu and focus on finding better ways to monetize the crown jewels of Sun; Java, Solaris, MySQL, along a few other ideas in the Sun R&D labs that have been hamstrung by being in a corporation that has major problems marketing and selling software for a profit.
"Except where they aren't and, er, never did, well not CPUs anyway... In fact unless I'm much mistaken Sun used to use Motorolla 68000 series chips back in the day (albeit the very sweetest batches only)"
Correct, but their performance sucked increasingly compared to, well, SPARC, MIPS, ARM...
"and, like apple, eventually ended up selling x86 boxes partly because it's widely regarded as NUTS to try and make your own processors these days."
Yes, ARM and all those licensees are insane, clearly. Not successful, either. Where's Matt Bryant when you need this level of argumentation?
You may like to consider that Larry has his sights set on ......... Dominating the Virtual Machinery SeeScape, ......the Visionary Field with Everything under Singular Control, if not necessarily All In House.
And that is a Field in which No One has Historical Advantage as it is Relative Brand Spanking NeuReal. And a Pioneering Field always Needs ITs Champions/Trail Blazers/Walkers on the Wild Side.:-)
I always worry when people listen too much to what people say. I suspect if we loose Sparc & Power, AMD won't be too far behind.
Also x86 architecture is not fit for every purpose under heaven. I'll bet that id Sun & IBM put their minds to it they can scale out their CPU's a lot faster and cheaper that Intel can, I'd also lay odds that Intel would tone down their R&D if they had only AMD for competition., so where as know we are getting faster and better CPUs at a clip, it would go down to walking pace.
Perhaps too many people have forgotten how IBM behaved in the early days, do we really want Intel behaving like that, and if you think they wouldn't you're even more NUTS that the people who buy Sparc servers.
Would you change your Mercedes Benz for a Proton, I know, I wouldn't. (Ok unfair, maybe a Ford then, think Via think Proton)
"Wait until Oracle gets the wakeup call of HW competitors such as IBM, HP, and Dell making a major push to recommend other software than what Oracle is selling to their customers. Why would they want to subsidize a competitor?"
They don't want to subsidize a competitor, but they don't really have a choice. Customers chose Oracle over SQL for a reason, and in this business climate they aren't about to pay for re-engineering their enterprise applications or pay for new training because of this move. It's a whole lot easier to plug in a new server (particularly one of Sun's x86 servers). Dell could be the primary victim of this move.
I don't think many people would go on record lying deliberately to the SEC.
Probably not even Larry.
"Wait until Oracle gets the wakeup call of HW competitors such as IBM, HP, and Dell making a major push to recommend other software than what Oracle is selling to their customers. Why would they want to subsidize a competitor?"
Right, because of course it is a well known fact that hardware manufacturers have any say in what software is deployed at customer sites!
"There is no way that SPARC innovation, based on limited production volumes can keep up with Intel/AMD R&D; they are being bypassed at a rapid pace by the competition."
If this were true why is it that Sun's new Rock processor will be the first commerical CPU to include hardware scouting and transactional memory? Intel and AMD may be adding more cores to their CPUs as fast as possible but this approach will very quickly run into diminishing returns. Adding more cores without other architectual innovations to enhance performance is a dead end.
The CPU mareket is changing quickly but those changes I believe ultimately work against Intel and AMD. The most significant change is that the needs of the home user and the data center are diverging quickly. Big multi-core CPUs may be fine in the data center but make no sense for the average PC user at home. Intel and and AMD have done very well because they were able to use essentially one type of CPU to service the needs of both customer types. Processors for the home market give lots of volume and the server market gives very nice margins. This simple truth is that for most people the current CPUs Intel and AMD sell are fast enough for the bulk of home uses. Intel is clearly concerned about low power Atom based systems cannibalising its more expensive products. Its possible to imagine a point in the not too distant future where due to energy considerations CPU performance plateaus for home PCs. Similar things have happened in lots of other industries as they mature. Jet airliners aren't much faster today then they were in the early 60's. Fuel efficiency and operating costs have dropped but flight times haven't changed much.
Oracle/Sun only has to concern itself with designing SPARC systems tuned to the requirements of the data center and particular workloads. I'm not saying that Intel doesn't have the resources to build separate products for each market because it clearly does. The Itanium family is an example of this. My point is that the server CPU market is changing and I believe that Oracle/Sun can market competitive products to exploit the opportunities these changes create.
Where's MB? He's bound to chime in with his trademarked FUD. Well, he claims it's not FUD, but pure unadulterated fact. That's what all FUDsters say though.
".....Where's Matt Bryant when you need this level of argumentation?" Patience, Grasshopper. The cunning fox waits for the bumbling Sunshiner to dig themselves a hole before piling the soil in on top of them. As Mr Heathcote correctly points out, Apple don't make their own chips and never have, so Mr Ellison is obviously planning a mass blitz of fanboi type marketing whilst dumping SPARC in the background. Or maybe he was referring to the fact that Apple don't have an enterprise server business as to what he aspires to, Apple don't make enterprise offerings, and what few servers they do offer are using Intel's CPUs. Either way, it is bizarre for the CEO of an enterprise software company to want to make their new (and recently failed) enterprise hardware business into a consumer-orientated software and hardware bizz like Apple. SPARC is not a good fit for iBones or iPlods, and a CMT chip in a Mac? Yeah, right!
As for CISCO, they have spent years raking it in using over-priced networking kit against precious few competitors. That virtual monopoly has taken a big hit from cheap Chinese LAN switches and an even bigger hit from hp's ProCurve. Their first attempt to sow up SANs has been kicked in the pants by Brocade, and their second attempt with FCOE - which was supposed to leverage their installed LAN/WAN base to dominate SANs - looks to have just opened their accounts up to attack from Brocade leveraging their installed SAN base instead. CISCO are at the same point Sun were nine years ago - diversify or die! Sun didn't diversify, they believed their own hype and look what happened. The big cash lump in the bank won't keep CISCO going forever at the rate they are losing marketshare, hence the move into servers, though that is a risky choice given it is such a competitive arena.
Bringing ARM or MIPS into the argument is moot - Oracle don't make those either.
RE: The hardware business
"They don't want to subsidize a competitor, but they don't really have a choice....." Actually, it is Oracle that don't have the choice. Sun makes a tiny minority of the server hardware that is used to host new Oracle apps. There may be a mass of old SPARC kit running old Oracle DBs, but those are ripe targets for hp and IBM Xeon kit running Linux and Windows. Looking forward, for Oracle to continue making money they need as many new installations as possible, and that means being nice to hp and IBM as they own the lion's share of the server bizz. There is nothing to stop hp and IBM simply putting more money and effort into competitor solutions such as DB2 and Websphere, or JBOSS and MariaDB, etc, etc. When hp and IBM were planning to announce full support for Linux back in the nineties, many said that Microsoft would punish them. As it turned out, M$ needed them just as badly, and the result is that hp and IBM are both leading Linux and leading M$ partners. Sun railed against both M$ and Linux and simply died.
"....Customers chose Oracle over SQL for a reason...." Yes, as a user of large Oracle RAC instances I can agree there are times we would only use Oracle. But nowadays it is mainly the enterprise high end, and M$ SQL has already surpassed Oracle DB as the most common database in our business and in the market as a whole. We use M$ SQL widely in what could be referred to as the second-tier and departmental environment. M$ SQL is getting more capable with every release, and there are instances where we use SQL Server 2008 where we would previosuly only have considered Oracle. As regards the rest of the Oracle stack, more and more of that business is being eaten up by opensource's offerings or Microsoft's. I have previously wondered how Oracle could diversify, and buying a failed hardware bizz does not seem the best of ideas. And I suspect Larry secretly agrees - first time round he asked hp to buy that side of Sun.
Personally, I find the whole idea of Larry bigging up the Sun hardware bizz quite bizarre, and if I was an Oracle shareholder I would consider it alarming! It was largely the hardware bizz that killed Sun, simply repeating the same mistakes with added Oracle licences is not likely to appeal to those shareholders. The line about SPARC/Slowaris being cheaper to own is just laughable. For a start, those Sunshiner admins price themselves rediculously high compared to your average M$ admin; the Sun management options are a joke compared to Windows; and then there is the fact the hardware costs ten times as much to give the same level of performance with the business apps people actaully want to run. And let's not forget Linux, which has been cheerfully gutting the SPARC/Slowaris bizz for years by being both cheaper, faster and easier to manage than Sun's offerings. Maybe he's hoping all the waffle will make IBM, hp or Fujitsu put in a bigger offer for the dross he needs to jettison.
/I never thought I'd say it, but now I'm pointing and laughing at Larry.
I'm amused at the comments about people being nuts buying Sparc servers and so forth, I've seen apps benchmarked inhouse where a T5240 beat a HP quad socket AMD box, thats just the reality. I'd say people are nuts for assuming the marketing blurb is real and not actually seeing how individual apps run in-house.
Assuming Oracle/Sun keep kicking out more designs such as a KT & maybe Rock then that keeps competition alive and I would expect Oracle to bang in money to accelerate the design process's and so forth. Sun is good on R&D but could do with shrinking the portfolio a bit, their R&D could do with being a bit more concentrated & faster.
Add in an Oracle licencing model change from per core to per socket and you'd be needing to reverse your statement.
Consider power+cooling, performance benchmarks, layered software licence costs, OS reliability + OS is free & maybe you'd be considered nuts for buying x86+M$/Linux or worse, Itanium +HPUX (factor in a cost for the OS as well).
The M-series stuff is fantastic as well if you need hardware reliability, CPU Instruction retry & mirrored memory banks, combined with Solaris self healing & thats getting reliabilty normally seen in mainframe systems. Hence I wouldn't view Sparc as something to be subsidized or be nuts over, they'll make more from it than Sun managed while independent just through increased volume.
I wanted to believe this was going to end up being a good move for Sun. However I'm getting the distinct feeling that a couple years down the line when king larry is saying "well we tried" as the last vestiges of Sun that Oracle couldn't immediately monetize are sold off.
pure & simple snakeoil - if this guy told me fire was hot, I'd have to check... don't trust hime one little bit
Hang on, isn't this is the same guy (and company) that's championed every year since 1988 as the year of the thin client? I've lost count of the number of times ORacle and Ellison have tried to convince me and the rest of the industry that they a) know what they're talking about and b) this is the year of the thin client - really, this time we mean it.
I'll believe that they'll make the SPARC a success after I start believing that this is the year of the diskless workstation/thin client.
"I don't think many people would go on record lying deliberately to the SEC.
Probably not even Larry."
Larry isn't lying to the SEC; he could simply get get a "better" offer than he expects to get for the HW division, give it a try for a quarter and declare it a failure, or simply state that market conditions have changed.
It is naive to think that what Oracle tells the SEC today locks them into any business decision a few weeks or months from now.
This was a Q&A released by the Oracle PR department thru emails to a pawn reporter at Reuters.
If you know Larry you know those are not his words. And who ever heard of an "email interview" and and "email interview" that has follow on questions? It's an embarrassment they try to pass that off as journalism.
Oracle said in their acquisition meeting they will make $1.5B in profit the first year. The only way to do that is fire 50% of Sun's workforce and raise the prices of Sun products 50% (just like they did with WebLogic last year).
ROCK is dead. Oracle is not stupid and they want less products not more.
M-class is going to see HUGE Solaris price increases in license and maintenance.
Fujitsu will see all of its profit moved to Oracles software stack.
T-class will survive and Oracle will try to move all WebLogic customers to Oracle x86 or NiagaraIII when that comes out.
Exadata is dead. But then again Oracle has not sold a single box. It will be replaced with Oracle first "appliance" and based on Sun x86 systems and storage.
Tape? Keep that? are you kidding me? oh thats right they could be sued by Sun if the deal does not go thru and they hurt Sun's tape business but saying they will kill it.
Fujitsu just sold their chip fabrication business and moved to Taiwan TSMC they see the writing on the all that SPARC64 is dead and Oracle will steal all their profits because they need Solaris. STOP....Do not buy an M5000!!!
Unfortunately, I don't see keeping Oracle on HP-UX. Oracle has never liked Itanium and still refuses to put the eBusiness suite on Itanium. This is a sad day for HP-UX customers.
Matt Bryant Rules.....Sun drewls
..... By Matt Bryant Posted Friday 8th May 2009 16:11 GMT
Larry isn't Pottering around in Hardware/Software fields, he's Flying High, Wwwide and Handsome in Vertical Virtual Cloud Infrastructure ... although he would most probably prefer the Analogy to be Sailing with the Breeze and Running before AI Perfect Storms. :-)
You've Misunderestimated the Situation and the Industry Position and Misinterpreted the MetaDataBase, although that is QuITe Understandable given the Fundamental Change Paradigm which is Seeing Profits Decimated and Wiped Out to be Replaced by Intrinsic Value and Flash Credit and Money for Novel Spending rather than Old School Banking..... which Kills the Golden Goose Gene/Midas Meme.
But no Matter what Regimen you Believe in, it will be in EVERY CASE, and I make no Apology for SHOUTING, Always an Intelligence Led Operation........ with the Best Intelligence Service being the Most Widely Read and Shared for a Greater Mutual Intelligence, which would then Really be, an Artificial Amalgam and a QuITe`Alien Construct in Words that can be easily Parsed for CyberIntelAIgent Use. ........... in a Sublime Human Readable CodeXXXX for Virtual AIMachinery.
Or would you Care to Disagree, rather than Agree to Disagree when in Doubt?
".....I've seen apps benchmarked inhouse where a T5240 beat a HP quad socket AMD box,...." Whilst I'm sure there are situations, such as skinny-thread webserving, where a T5240 might have an edge over a four-socket Opteron box in raw performance, I'm guessing price didn't form part of that discussion seeing as the hp would be much cheaper. Probably so much cheaper you could get two of the hp servers for the price of the Sun solution. The market obviously thinks so as just the number of hp Opteron servers going out each quarter is higher than the Niagara lines. And that's because for the majority of customers, seeing as they don't want Slowaris, the T5420 is completely irrellevant as I can't run the Windows or Linux apps they actually want to run.
OK, let's just give the Sunshiners a chance - can one Sunshiner post exactly why it is that even though Slowaris is supposedly so great, SPARC (including Niagara) just so powerful, and Sun so good at R&D, why is it that Sun has crashed from a market-dominating $200bn company to a failure bought up for the tier 1 equivalent of chump change? Then how they really expect Larry to execute the same failed startegies and somehow make them a success? Please try and keep at least a tiny portion of reality in your responses, including why you think the market is so clearly moving away from Slowaris and SPARC to Windows and Linux on x64?
/not holding my breath for a non-fantasy response - well, hard to when you're laughing this hard!
It's not like Oracle hasn't owned hardware vendors before - their experience with nCUBE (1988-2002) was highly interesting, and really helped Oracle mature their parallel database thinking in the early days. I did a number of benchmarks on the nCUBE at the Northeast Parallel Architecture Center in Rochester in '93, and it was a decent enough machine.
I think the important consideration is SYNERGY, between the hardware design and software architecture. Database software is able to use very high degrees of parallelism in the CPU, microarchitecture, and overall systems design - probably more so than any other type of software. Sun had some very talented people working on highly parallel architectures that are a natural fit for that, and Oracle is and always has been the best parallel database (besides specialist Teradata). I think there may be some synergies there, and I would expect that Oracle and Sun may be able to get some very nice results working together closely. In someways this is a return to the olden-days of parallel processing, where many database vendors had proprietary hardware systems (Teradata and the Y-network, Oracle and nCUBE, MassPar and Transputers, etc.). Ahh, those were the days...
Oracle has the cash, and Larry is a risk taker - and part of me thinks he just likes the challenge of saving Sun or extracting what value lies within. Good on him if he makes it work, as it can only drive the market forward...
Obviously no one bothered to read the whopping disclaimer at the end of the PDF: "Cautionary statement regarding forward looking statements".
My favorite bits are:
"Any such statement may be influenced by a variety of factors, many of which are beyond the control of Oracle or Sun, that could cause actual outcomes and results to be materially different from those projected, described, expressed or implied in this document due to a number of risks and uncertainties"
"Accordingly, no assurances can be given that any of the events anticipated by the forward-looking statements will transpire or occur, or if any of them do so, what impact they will have on the results of operations or financial condition of Oracle or Sun."
There's more. What it basically means is that Oracle reserve the right to do a U-turn on anything LE has said at any point in time if they think it makes business sense to do so. In fact the disclaimer also says:
"You are cautioned to not place undue reliance on forward-looking statements, which speak only as of the date of this document."
So buyer beware.
Personally I think that Larry is smoking something that's clouding his judgment. He seems to be overlooking one important fact about Apple and Cisco. In both cases the synergy they create with the hardware and software are something only they enjoy. Apple for example do not sell Mac OS X or the iPhone software to others to put on 3rd party hardware. Apple are the only source of OS X and iPhone software. That is not the business model that LE is proposing.
Article say, "Oracle would stick with the SPARC architecture, he said, pointing out that "even Apple is designing its own chips these days".
Roger Heathcote posts, "Except where they aren't and, er, never did, well not CPUs anyway..."
Apple purchased P.A. Semiconductor and has been hiring chip design expertise.
Roger Heathcote posts, "In fact unless I'm much mistaken Sun used to use Motorolla 68000 series chips back in the day (albeit the very sweetest batches only)"
Back in the days of segmented intel architecture, programming was a proverbial nightmare - the 68K architecture offered a reasonable model. The entire industry (not just Sun and Apple) produced servers, workstations, and desktops on 68K.
Roger Heathcote posts, "and, like apple, eventually ended up selling x86 boxes partly because it's widely regarded as NUTS to try and make your own processors these days."
There was no comparison between the 16 bit 8086 and the 32 bit 68000 processors. It was not until Intel created the i386 with a flat memory model capability that Intel become competitive, offering similar CPU architecture to Motorola. At this point, other vendors experimented with migration due to availability and later performance.
Roger Heathcote posts, "I'm sure Oracle love the thought of the lock in that would get them though!"
SPARC is the wrong architecture or Oracle to buy for lock in. SPARC is completely open specification, now open-source, and multiple companies have always created competing lines of SPARC chips (without worry about lawsuits) - which means there is nothing for Oracle to lock in.
The Oracle takeover of PeopleSoft doesn't inspire confdence. But whatever your attitude towards Oracle and Larry Ellison might be, all platforms other than x86 - PowerPC, SPARC, Intel's own Itanium - face a very difficult future in competition with the enormous economies of scale that the x86 is blessed with.
At the very low end, where it is hard for a full-fledged x86 processor to compete, although the Atom is now making inroads, things are different, at least for the moment.
I wish there were other choices besides x86, but for a choice to exist, those behind it will have to be very realistic.
Oracle operates at a 70% margin, Sun a mere 40% ... no go where the money is :
Cheap CPUs in generic Boxes (HP Dell ... whatever),
2006 bechmarks (Q4) SPECfp_rate_base2006
Sun SPARC Enterprise T5440 Server
(benchmark = 212, discounted price: $31,077.00)
Dell Inc. PowerEdge M600 (Intel Xeon E5450, 3.00 GHz)
(benchmark = 71.7, price $3,711.00)
that makes the Sun's benchmark point 3 times more expensive than the Dell's one.
Now call me stoopid but in the current downturn SPARC is dead as a can of spam.
/need a $ icon.
Larry is so far up..( actually I'll use the polite version! ), so deep in Steve Jobs' pocket, this is hardly surprising at all!
"SPARC buy means we're just like Apple"
Just what we need, another Apple ;)
WinHatter posts, "that makes the Sun's benchmark point 3 times more expensive than the Dell's one. Now call me stoopid..."
One gets a lot more than just CPU integer and floating point capacity when they get an T2 processor from Sun. There is integrated dual 10 GigE and octal crypto acceleration in the same chip. You have to look at the whole package.
If you look at the SPECweb2005 benchmarks, a 1 socket T2 scores ~41.8K while a 4 socket hex-core intel scores ~51.3K. The pricing is very competitive.
While the T2 processor can scale to 4 sockets in a single chassis, the Intel hex-core will not scale to 16 sockets in a single chassis.
When the pricing and performance is competitive, but you can scale more CPU's into a single (UltraSPARC T) chassis, and those chassis can save space (as well as power and HVAC costs) to deliver the same workload - one can see the benefits.
More Novatose hilarity!
"....One gets a lot more than just CPU integer and floating point capacity when they get an T2 processor from Sun....." Yeah, you get over-priced and inflexible support, a lack of applications and no backward compatibility with your old SPARC servers. Compare to x64, which is backwards compatible with years of Intel x86 code, where competition has honed a choice of vendors with polished support at a good price, and a wealth of apps on Windows and Linux (and even Slowaris x86).
"....There is integrated dual 10 GigE and octal crypto acceleration in the same chip...." And which can be added into a base Xeon by generic cards for a fraction of the cost, that's if the customer even wants them. Market figures say most customers don't.
"....You have to look at the whole package....." The problem for Larry is the market did look at the whole package and didn't buy Sun's kit.
"....The pricing is very competitive...." Have you actually looked just at the tin prices, let alone all the other costs? Sun haven't been price competitive for so long it's beyond funny.
"....While the T2 processor can scale to 4 sockets in a single chassis, the Intel hex-core will not scale to 16 sockets in a single chassis...." So you admit the Niagaras can't scale as high as Xeon but that's an advantage? Or are you counting wheiner threads which can't all work in parallel as compared to Xeon threads that can? Oh, by the way, Unisys E7000s have scaled to 32 Xeon sockets since 1999, which kinda makes you look twice as stoopid.
/Novatose - the gift that just keeps on giving!
Matt, one simple answer to your request for a comment on why Sun marketshare has dropped : Oracle, the largest Unix related ISV (Matt, stop frothing, I know they make software for non-Unix OS's you pedant!) had priced it's software (via CPU core multipliers, .75 per core on T2 instead of .25 per core on T1) and that makes T2 Sparc server equipment less attractive pricewise even if performance is cracking and they hold bucket load of benchmarks. Hence Oracle DB instances on Sun have dropped. Whether you'll admit it or not the largest source of Server sales for Unix is for corporate databases.
However, you may have noticed who just purchased Sun hence this thread? I'm not sure it's sunk in yet unless your in denial of that little nugget.
Couple it with the Solaris x86 debacle a few years ago plus general Linux/M$ reliability improvements and you can see where other market share went. Note : Oracle cited Solaris as the preferred OS so looks like thats in the air again. x86 linux "cheap and not quite good enough" slowly became "cheap and just about good enough" but then Sun/Oracle get serious with Solaris x86. Ent. grade Unix on x86 with lower support costs? Hmmm......
Anyway, gotta dash, expecting another thrilling installment of HP drivel from you, due any moment soon! (keep it under 1000 characters this time, each ascii character has carbon content don't you know.)
I think we're probably screwed the amount of hot-air you create!
DavidHalko posts, "One gets a lot more than just CPU integer and floating point capacity when they get an T2 processor from Sun"
Matt Bryant posts, "over-priced and inflexible support"
Incorrect, you can get support from many different vendors.
Matt Bryant posts, "no backward compatibility with your old SPARC servers"
Incorrect, application binaries run transparently, I have not found one.
Matt Bryant posts, "Compare to x64, which is backwards compatible with years of Intel x86 code"
DavidHalko posts, "There is integrated dual 10 GigE and octal crypto acceleration in the same chip"
Matt Bryant posts, "which can be added into a base Xeon by generic cards for a fraction of the cost"
At a loss of energy efficiency, system cost, rack density, reliability (with more components)
DavidHalko posts, "The pricing is very competitive"
Matt Bryant posts, "Have you actually looked just at the tin prices, let alone all the other costs?"
On the quad socket hex-core box - I did not add the costs of the nearly dozen nic cards used in the benchmark by the Intel box, the increased number of network ports by the Intel box, the increased energy usage by the Intel box. I was very generous to the competing platform.
matt bryant ~ Unisys E7000s have scaled to 32 Xeon sockets since 1999, which kinda makes you look twice as stoopid.
you'll need a room full of those 1999 servers with old xeons - better post a link to that modern unisys e7000 with 32 xeon hex-core sockets, stupid!
ha ha ha ha ha!
"Matt, one simple answer to your request for a comment on why Sun marketshare has dropped : Oracle, the largest Unix related ISV (Matt, stop frothing, I know they make software for non-Unix OS's you pedant!) had priced it's software (via CPU core multipliers, .75 per core on T2 instead of .25 per core on T1) and that makes T2 Sparc server equipment less attractive pricewise even if performance is cracking and they hold bucket load of benchmarks. Hence Oracle DB instances on Sun have dropped. Whether you'll admit it or not the largest source of Server sales for Unix is for corporate databases....." Yes, but by your argument, only Niagara would be affected by Oracle's pricing startegy, so how do you explain the continuing nose-dive in other SPARC lines? The obvious answer is that the fact the kit came from Sun (or FSC) made it less attractive than other vendor's (IBM and hp) kit. The real reason? IBM and hp provided UNIX kit that not only did the job but did it better, cheaper and with more of a sense of security than Sun could provide. It's the market that choses winners and Sun has been marked the loser by their defecting customers.
"....However, you may have noticed who just purchased Sun hence this thread? I'm not sure it's sunk in yet unless your in denial of that little nugget...." Yes, Oracle bought Sun. Did you also notice the subject of the article? Oracle didn't originally want the Sun hardware bizz, they asked hp to take it. Oracle are a software business, no hardware. You mentioned denial, but you're shouting from ten feet underwater in a certain Egyptian river.
"....Couple it with the Solaris x86 debacle a few years ago plus general Linux/M$ reliability improvements and you can see where other market share went. Note : Oracle cited Solaris as the preferred OS so looks like thats in the air again. x86 linux "cheap and not quite good enough" slowly became "cheap and just about good enough" but then Sun/Oracle get serious with Solaris x86. Ent. grade Unix on x86 with lower support costs? Hmmm......" And seeing as hp, Dell and IBM can offer that Slowaris x86 support, still no certain revenue stream there. Of course, the fact that Sun's own Galaxy customers ordered Linux as the OS of choice over Slowaris by a ratio of 5:1 kinda shows the market doesn't belief Slowaris is the OS of choice. Enterprise Linux such as SuSE and RedHat is not only more popular, more mature and has a bigger community than Slowaris x86, it also has many years' headstart and far stronger backing from the real power x64 vendors (that would be hp, IBM and Dell, not Sun).
".....Anyway, gotta dash, expecting another thrilling installment of HP drivel from you, due any moment soon! (keep it under 1000 characters this time, each ascii character has carbon content don't you know.)....." Well, you stop repeating the same Sunshine every post and I won't feel the need to constantly remind you of reality.
"....I think we're probably screwed the amount of hot-air you create!" Please try reading without moving your lips then, that should cut down the hot air a tad or so.
/SL&P, but need new Sunshiners as the old ones are just getting boring.
And Novatose manages to completely miss the reality again. It's quite worrying that he actually seems to believe what he posts.
".....Incorrect, you can get support from many different vendors....." The fact the market is not buying Sun shows you those options are not attractive enough. OpenSlowaris/Slowaris x86 is not generating cash for Sun and won't for Oracle. The majority of installs are not on Sun hardware so even if the users do decide to pay for support, the majority will likely take advnatge of bundled support from the hardware vendors, which means the majority of revenue will not go to Oracle. So-called OpenSolaris unchanged will be just as big a profits blackhole for Oracle as it was for Sun.
"....Incorrect, application binaries run transparently, I have not found one....." I'm assuming that's just poor English or a typo, and you meant you had not found any incompatibilities. The biggest incompatibility is performance. When we did migrate our old Slowaris binaries, the easiest route was to shift them to Transitive on top of Red Hat Linux on Itanium - no need for the porting exercise Sun suggested to shift them to Niagara Slowaris 10, and actually performed better than either the original Sun servers or the Niagaras Sun wanted us to buy. The main reason was the old SPARC apps were heavy-threaded and included Oracle - performance on Niagara kit was dismal compared to even the old SPARC kit, and then there was the licensing cost of Oracle on Niagara! And guess what - the majority of the aged Slowaris installed base is in exactly the same positiion as we were - heavy-threaded apps that won't perform on Niagara, and many with Oracke DB. They are going to be looking at hp-ux and AIX, seeing an easy migration and better heavy-threaded capability, and the uncertainty surrounding the Sun hardware bizz. Doesn't take any crystalball gazing to predict that hp and IBM will be cleaning up those Slowaris accounts, either with AIX and hp-ux, or just Linux on Xeon/Opteron.
"....On the quad socket hex-core box - I did not add the costs of the nearly dozen nic cards used in the benchmark by the Intel box, the increased number of network ports by the Intel box, the increased energy usage by the Intel box. I was very generous to the competing platform....." Unfortunately for you Sunshiners, what has made it easy for hp and IBM is that when customers do the same pricing exercises they are finding xSeries and ProLiant with RedHat or SuSE are far cheaper. Even M$ Windows Server charging more than Linux, it is still eating up the Slowaris base! And Niagara can't run WIndows, RedHat or SuSE, so it offers the least flexibility along with the highest costs. If anything you said was relevant then the opposite would be true and Slowaris on Niagara would be increasing Sun marketshare in all segments - it's not, Sun has been declining for years, and Niagara has not made any change in the steepness of that decline.
"....you'll need a room full of those 1999 servers with old xeons - better post a link to that modern unisys e7000 with 32 xeon hex-core sockets, stupid!...." Ignoring the fact that Novatose said Xeon didn't scale, which the E7000 example disproves, please go look at the Intel chip sales figures, look at Sun's, then try carrying on laughing. Better still, look at the profit figures for Intel, hp or IBM, then go look at Sun's. Still laughing? I doubt it, especially if you are a Sun shareholder. Only last year, The Reg posted an article on how SGI and Unisys were canaries in the coal mine, and people like Sun needed to take note to avoid dying. Well, McNeedy and Ponytail had taken note - they put Sun up for sale. All your Sunshiner bluster is revealed by the simple fact that Sun's own management had no faith in their own company's ability to survive without a purchase. Continuing to post Sunshine now we all know that just shows how really, really stupid - or just plain gullible - you are.
And as to an Unisys E7000 with Nehalem Xeons, who knows? Last time I looked, Unisys were still (barely) alive and planning for the MS SQL 2008 R2 release. At the moment, given the complete lack of roadmaps released by Oracle, the E7000 seems to have a rosier future than any Sun server product. Every time a server vendor dies it offers a chance to the other vendors to feast on the dead company's installed base, and at the moment I can bet there are people at Unisys planning on how they can attack all those dying Sun SPARC accounts and maybe steal something away from the bigger boys.
/the new AC Sunshiners are even stupider than the old ones!
Biting the hand that feeds IT © 1998–2017