back to article Ellison: 'There'll be nothing left of IBM once I'm done'

Oracle has pulled the rug out from under Hewlett-Packard's Intel's Itanium processor by yanking support of its database, middleware, and application software on future "Poulson" and "Kittson" Itaniums. It looks as though Larry Ellison wants to take on IBM in microprocessors for data center systems, man-to-man, head-to-head. "I …

COMMENTS

This topic is closed for new posts.
Silver badge

@kebabbert

"Again, if you are doing it right, you never use floating numbers in finance. Every calculation is done with integers, and you keep track of the number of decimals separately. No rounding will occur. No floating numbers are needed. As I said, I work in a large finance company."

Hmm, well I'm not sure that you've wholly understood what that part of POWER is. As you say, floating point is no good. I gather that the *decimal* accelerator (not the FPU) in POWER is doing the sort of arbitary precision sums you've outlined. Hardware acceleration of that is obviously going to bring benefits, or so IBM would have you believe. And it's difficult to disagree with the evidence of their sales figures.

"Those are trivial calculations, done in COBOL on Mainframes. Not very sexy."

Not sexy, but clearly very profitable. Profits don't have to be earnt in a sexy way, they just have to be big! I'd say that on an absolute scale skyscrapers, suits and MBAs are not really significantly sexier...

Yes, I'm familiar with the technology that the high speed trading world uses, and I'd certainly agree with you on the inappropriateness of mainframes in that role! But generally I think that Solaris/Linux on even top end server and network hardware is behind the curve when it comes to low latency, largely because they're stuck with stodgy sluggardly interconnects like Ethernet, Myrinet and Infiniband.

The high performance embedded signal processing world has been much more focused on low latencies than the mainstream server world. The 'unconventional' interconnects found in that domain (VXS's and OpenVPX's sRIO, and external interconnects like sFPDP) are all about low latency. That's because it's a key driver in the sorts of applications (radar, etc) implemented using such hardware. If you've not done so already, it's worth a bit of investigation.

Anyway, hurry up and earn your profits. There's a good chance that the whole high speed trading thing will get banned soon, especially if it gets fingered for causing a major market wobbly. I'm pretty sure that no one in the finance industry could say whether or not it meets the Nyquist stability criterion, but those of us who know what that means generally think that it doesn't and don't believe that anyone's checked to see either.

Even if it doesn't fall over it's doomed stagnate, eventually. As soon as you've all bought premises as close as possible to the stock exchange and have all chosen the optimum hardware and algorithms for the job, you'll all be as good at it as each other at it and there won't be any technological advantage left to exploit. Anyone started checking to see if they're plugged in to port number 1 on the Exchange's Ethernet switch?

On the otherhand if stagnation spurs the finance industry into developing even lower latency kit (the mainstream IT industry won't, they care merely about throughput) then I would be quite grateful.

3
0
Anonymous Coward

Citibank

There is a reason Citi, BofA, JPMC and Wells have standardized on IBM Power and a SPARC box requires 3 levels of escalation.

2
0
Alert

FORTH now!!

Back to FORTH!!

0
0

@bazza

Regarding integer calculations or floating point calculation - I dont understand why you IBMers are talking about it so much. The problem of doing such calculation is not a very big problem. The vast majority of server cpus have no problems of doing such calculations, they have no performance problems. The problem is fetching data. Of course, if you are doing BIG calculations then it is a problem. For instance, to multiply really really huge numbers takes long time, the best algorithm are using FFT and does that i O(n log log n). But I dont think such calculations are done. Only trivial calculations are done, such calculations no cpu have problems to do.

Regarding embedded signal processing, I will check that up. Thank you for that. Our system is among the fastest in the world today, and if we could lower latency even more, it would be even better. But I have doubts those solutions transfer directly to servers and IT.

1
6

@A.C

Ok, I did not know that. In all my travels around the world, I have never seen AIX nor POWER. Do you what these banks are using the IBM systems for?

0
9
Anonymous Coward

What do you mean "back to"? I still use it most weeks.

1
0
Headmaster

@Kebabbert @bazza

I think you'll find that if you are talking about mainframes, transaction values and bank balances that's it's all stored and calculated in packed decimal format, the mainframe CPU uses a complex instruction set that has machine code instruction for dealing with this rather archaic format.

“Mainframes are never used - they are too slow. Here we typically use Linux/Solaris and C++ and do High Frequency Trading, algo trading, quant math, risk analysis etc - now THAT is sexy! Banking world is boring. Finance is cool. I see sky scrapers, suits, MBAs, Quants, Traders, HFT, algorithms, Hedge Funds, Wall Street, etc”

If my memory serves correctly there was a least one financial crash caused by your fast, sexy, computers, where the computers got stuck in a negative feedback loop and started selling anything that wasn’t nailed down, that one comment alone warrants a FAIL icon. Finance is cool, only if you think loosing money is cool.

“To err is to be human, but to really fuck things up you need a computer”, especially if you are using the computer to perform quant math, risk analysis etc. for credit default swaps, derivatives, hedge funds and naked short selling and a mess of other shit that has fucked up the worlds economies.

2
0
fch

Both Power and SPARC servers have Forth built-in

... and they don't even need an operating system for it. They both use OpenFirmware, and the "ok" prompt there will give you a full-featured Forth interpreter right at your fingertips.

0
0
FAIL

Get out much?

"In all my travels around the world, I have never seen AIX nor POWER." Looks like you have missed the boat. IBM keeps showing me these market share numbers that shows AIX and Power are now over 50% of the Unix market

"Do you what these" Hopefully, English is not your first language.

Shouldn't you be at OOW with the rest of us?

1
0
Silver badge
Stop

Larry forgot his own memo!

After all, it was Larry that first tried to get hp to buy the Sun server bizz, then tried to flog it to any taker. Larry is so full of it, it's hardly surprising he's misplaced a few memos amongst the brwn stuff.

http://www.theregister.co.uk/2009/06/19/oracle_shopping_sun_hardware/

http://www.theregister.co.uk/2009/05/12/sun_oracle_sec/

2
3
Silver badge

IBM, now there is someone with a massive pile of patents

I wonder how long it will be before they start playing at being like Apple and going after every little thing Larry wants to do.

1
1

@Dazed...

Just like when IBM nearly bankrupted Sun? As Gosling explains:

http://nighthacks.com/roller/jag/entry/quite_the_firestorm

Or like this time, when IBM demanded money from Sun, saying "ok, you are not violating these 7 patents, but we have many more. Do you want us to come back with patents you actually do violate or are you going to pay?". Sun payed again.

http://www.forbes.com/asap/2002/0624/044.html

I would not cry if Oracle crushed IBM, for obvious reasons. Larry is the Man. :o)

2
12
FAIL

You have Ellison Envy

That's worse then Penis Envy.

The difference between God and Larry Ellison? God doesn't think he is Larry.

5
1

@Allison Park

So have you read the links? When IBM almost black mail other companies for money? You think that is ethical and fair, right? And when IBM is using foul play, you have no understanding that people react, and dont like the foul play by IBM?

0
7

By clear admission of denial you admit it

You obviously have Ellison Envy....

0
0

@Kebabbert

Great links, read both.

0
0
IT Angle

now we know why

now we know why you have a hard on for IBM.

Be it unsubstantiated claims with little regard for the facts it does make for interesting reading.

I would have thought that inventing RISC would be more than "make it simple and it will go fast", but I dont have time to troll the patent database.

Try to come up with articles from the last 3 years not 10 please

1
0

@Allison Park

These patent stories are only some of the foul play and FUD that IBM uses. The worst thing is that people seems to believe that IBM is not Evil? That is funny. But on the other hand, IBM marketing division is very strong, they boasts all sorts of strange things. For instance, that the z196 Mainframe is the "worlds fastest cpu" when in fact it is much slower than a decent x86 server cpu.

http://www-03.ibm.com/press/us/en/pressrelease/32414.wss

If one z196 Mainframe cpu at 5.26 GHz and 300MB cpu cache is slower than a x86 cpu, how can a big IBM Mainframe with 24 cpus virtualize 1.500 x86 servers as IBM claims? That must be a gross.. something.

.

Regarding my Larry Ellison envy, hell yea I am envious of him. To be one of the richest man on earth triggers my envy, yes. I am but a man. Not like you, that have all the money you need.

.

Regarding the RISC patent. RISC cpus are built like that. Instead of having complext microcode, you only have a few cpu operations, that are heavily optimized and only use them. Basically you throw out all unnecessary. Keep it lean n mean. That is faster than bloat.

One of the links that you think are Trolling, is written by James Gosling, father of Java. I can assure you that he is more credible than you, a random FUDer that never backs any of your claims up. At least I give links that shows I speak true.

The other Forbes link you dispute, is by

"Gary L. Reback has been named one of the "100 Most Influential Lawyers in America" by the National Law Journal"

and I think he also is more credible than you, a well known FUDer and Troll.

So, instead of accusing me bringing up old articles, how about you bringing up ANY article that supports your unsubstantiated claims? Oh, there are no such links, because you made up everything? Ok, I get it. Then how about you stop making up things in that case? Only write things you can support, either by a stringent and correct logical reasoning, or by supporting links. If you speculate, be clear with that and dont try to make everyone believe your speculation is true. I speculate, but then I am clear with that - I dont try to fool everyone my speculation is true.

1
7
FAIL

@Kebb

"Again, if you are doing it right, you never use floating numbers in finance. Every calculation is done with integers, and you keep track of the number of decimals separately."

Wrong. Read about Decimal Floating Point and there is a chance you'll understand. Available only in POWER and Z.

4
0
Silver badge
Thumb Up

@tom 99

Seems you're a deserved member of the group of a select few people I know who have properly read up on modern CPU technology.

IBM are in many ways infuriating. They're always coming up with neat tricks like this which have applications way beyond IBM's core business, but the only way you can get it is to buy a bloody banking system off them. It is admirable how they stick to what they do best and not waste shareholders' profits on supporting the few small fry like me who'd like to use their tech for something other than an entire country's credit card transactions processing...

2
0

@Tom 99

Where am I wrong? So when I say that our large finance systems that you have heard of (and probably used), are keeping track of decimals separately and never do any rounding - I am wrong?

What should I read about decimal floating point? Can you give me a link "so that I understand"?

Again, such trivial calculations are easy to do, and no cpu have problems doing that. The problem is fetching data quick enough to feed the cpu continuously. I dont understand why are talking about that IBM feature, it sounds as if it were as good as sliced bread?

0
8
Silver badge

BCD and Decimal Floating point are a waste of space

I have read about decimal floating point, but have never used it because fixed point integer has always been a better solution. I have used fixed point integer on 8041, 6502, Z80, 68000, TMS320, ARM, MIPS, X86 and AMD64. When I needed arbitrary precision, I used GMP or Python. If POWER or Z ever fit into the available budget, space and power requirements, I will stick with fixed point integer.

0
0
Silver badge

BCD is only an IBM thing

I started my programming career working on 8bit micros that did BCD for their FP, that's what happens when the CPU is designed by the calculator division.

0
0

A lot of Java

"We think [...] run a lot of Java."

Yes they do, but

-for how long? Ellison is doing a fine job at driving people away from Java. Red Hat and Google are not to be taken lightly these days...

-they are also running a LOT of COBOL (probably much more than Java, in terms of lines of code).

"and then there will be nothing left"

There will be the reputation for reliability and dependability. What Oracle has now is the reputation to shaft their customers and business partners at the first opportunity.

Also, IBM "has" currently 212 systems in the top500 (yes, really), Oracle "has" 12.

That's hardly "nothing left" when it comes to making a purchase decision, even if Oracle did in fact manage to build a benchmark system able to blast IBM in integer arithmetics.

1
0
Silver badge

K Machine?

"Also, IBM "has" currently 212 systems in the top500 (yes, really), Oracle "has" 12."

Maybe, but the fastest of them all is SPARC based. Half million+ SPARC cores and counting...

Ok, so Fujitsu makes the chips, but it's still good PR for Oracle and the SPARC commnunity.

0
0

Re: K Machine

Well, if you want to go that way, there are a grand 2 SPARCs in the Top500... 45 Powers. And I bet all 45 Powers are IBM's, while none of the SPARCs are Oracle's.

"We didnt build it but we use a similar architecture for our processors" does not sound like terribly good PR.

Of course the proportions might change, Oracle are pretty new in that game; all I am saying is that it will take more than building one benchmark system faster at crunching integers to leave nothing to IBM. And behaving like total dickwads and liars, always ready to stab their partners and customers in the back, is not going to help PR.

7
0

@ElReq!....

The systems running COBOL are Mainframes. I dont know of any x86 / POWER / SPARC that are mainly deployed to run COBOL.

And theoretically, Java is faster than C++, because adaptive compilation. And for your information, several of the largest and fastest stock exchanges in the world, are done in Java.

Regarding top500, it has no relevance for the discussion. In place 7(?) we find Blue Gene, which is very fast. It has many Power cpus running at 750MHz. Do you claim that such cpus are among the fastest in the world? No. So what does Blue Gene tell you? A supercomputer has other needs and requirements for their cpu, needs that say nothing about how good or fast a cpu is.

1
7
Silver badge

re K machine

+++

"We didnt build it but we use a similar architecture for our processors" does not sound like terribly good PR.

+++

Well, its not bad publicity either, and it is a small attention grabber.

I completely agree that Oracle are quite capable of cooking up their own bad PR! Like them or loath them, SPARC getting improved is only a good thing, but I think we'd all prefer a more modest communicator of that news... And indeed it will take more than a single benchmark to out do IBM.

1
0
Silver badge
WTF?

@Kebabbert

"And theoretically, Java is faster than C++, because adaptive compilation."

Oh gawd, not this again. Yes , if you take a jar file and stick it on some random machine the jv, may in some rare circumstances produce a more efficient runtime than a statically compiled C++ binary thrown onto the same machine.

Newsflash - production machines are generally NOT an unknown quatity and any company with any brains doesn't throw over binaries from a dev or test box , it compiles C++ binaries specifically for the machine they're going to run on. In fact in our company we compile them ON the production box.

"And for your information, several of the largest and fastest stock exchanges in the world, are done in Java."

That'll be despite java rather than because of it. The amount of hardware thrown at java these days to get decent performance is getting ridiculous.

2
0
Silver badge
Boffin

RE: @ElReq!....

"The systems running COBOL are Mainframes...." Not so, unfortunately. The Y2Kgasm meant many companies looked at replacing old servers and systems with newer kit, but they mainly kept the old code with a few Y2K edits. There were several reasons for this, mainly because the packages in question were very large and complex and closely tailored to specific business tasks, and becuase many of the original COBOL was written without good documentation or comments in the code. Completely rewriting the COBOL apps in newer languages was seen as too much of a task, so we ended up shifting a lot of old COBOL programs off platforms like old mainframes and VAX onto AIX, hp-ux and even Slowaris servers. I know of several City COBOL applications currently running very complex business solutions that have their roots in COBOL code from the '70s! We even have a team looking at the idea of porting at least one COBOL app to OpenCOBOL, just to make it future-proof.

0
2

Elephant in the room is J2ME

Imagine you got a technology which has been somehow implemented on billion devices and smart phones took off, making people even non tech adults think about "running an app on phone" as an ordinary thing.

It has gps, 3d, voice sampling and whatever you like support, standard.

What happened to J2ME Mr. Ellison? Any ideas about how to wake up the technology, assure device makers, re-gain developers?

"but they don't install apps" is an old apology. Opera Mini? Snaptu? IM clients?

Better fix your own stuff before declaring war to a freaking database/server giant?

0
0
FAIL

@Kebabbert

"The systems running COBOL are Mainframes."

Do your homework Kebabbert

MicroFocus, ACUcobol, Veryant, 32-bit and 64-bit on *nix & windoze using x-windows and windoze API to create widows programs, callable from Java, and cobol is OO in the 2002 spec.

Fail, epic

2
0

One press release is enough

Only problem with truly amazing Power7 is, price and multiple vendor support.

What would happen to Sparc if IBM says "IBM and Fujitsu announces cross manufacturing and R&D agreement for Power8 processors"

Hell, what would happen if IBM dumps the DB2 prices (or even free) for Sparc systems?

0
0

@Field Marshal Von Krakenfart

Ok, I should have written "The systems running COBOL are mainly run on Mainframes".

0
5

@boltar

No, you are wrong. In THEORY, Java is faster than C++. It might not be the case today, but the day will come when adaptive optimization will be faster than static compilation.

.

"...That'll be despite java rather than because of it. The amount of hardware thrown at java these days to get decent performance is getting ridiculous..."

In one of fixed income exchanges in Wall Street I recently visited, we had a discussion about this. Developers there said that one study at... CalTech(?) showed that a algo trading system written in Java were fastest. They implemented in different languages an measure performance.

Regarding large stock exchange systems written in Java, I have some internal numbers, but can not disclose them. But if you look at the fastest systems in the world several of them are written in Java. London Stock Exchange, LSE, system is written in C++ and runs on Linux/Solaris. NASDAQ is written in Java and runs on Linux. Look at the production numbers and you will see that Java can give extreme performance and latency - if you know what you are doing. Skip the test numbers, instead look at production performance.

So, it seems that Java rivals C++. Of course, LSE's earlier system were written in C# and ran on Windows - and crashed a lot and costed 50 million Pounds. So LSE threw it out after 1-2 years and bought Linux and C++ system for another 30 million Pounds.

1
6
Silver badge

@Kebabbert

"No, you are wrong. In THEORY, Java is faster than C++. It might not be the case today, but the day will come when adaptive optimization will be faster than static compilation."

Compared to a binary optimised for the specific machine its going to run on? Sorry but I'll believe it when I see it.

"But if you look at the fastest systems in the world several of them are written in Java."

QuickBASIC would run fast on the systems these people use. And comparing different setups is pointless , there are hundreds of variables outside of the application program itself.

3
0

Re: what does Blue Gene tell you?

If you re-read the post you are answering to, you'll notice that I was talking reputation. What having built nearly half of the Top500 machines tells the world is that you are able to build Big Iron at a price people are willing to pay. And that you can support your customers. What I was saying is that it takes more than building one single benchmark machine to overcome the reputation gap. People look at more than just speed when they spend that kind of money on a system. "Will it perform as advertised, and is it still going to be supported in 6 month?" are also big questions; with IBM you can answer "almost certainly" to both, with Oracle the answers are more "it might if you're lucky".

3
0
Anonymous Coward

sad

Sounds like your employer wouldn't pay for your trip to OOW. ouch.

0
0

@Boltar

"...Compared to a binary optimised for the specific machine its going to run on? Sorry but I'll believe it when I see it...."

Maybe you have not studied comp sci and compiler theory? Then I suggest you do that.

.

.

"...QuickBASIC would run fast on the systems these people use...."

Actually, these stock exchanges run commodity x86 servers albeit beefed up, and Linux. Nothing fancy nor special here. Trust me on this, or go and check it up yourself. They dont use supercomputers. Typically they are using several dual socket Intel Xeon 6-core rack servers with 36GB RAM or something similar.

Of course QuickBasic would run fast on such a system, but so what? C++ and Java get equal speeds. As I said, I have some internal numbers, but can not disclose them. If Java were too slow, then no stock exchange would use Java. But they do. The trick is to never trigger the garbage collector in Java, but manage the memory yourself.

1
4
Silver badge

COBOL on non-mainframes

"I dont know of any x86 / POWER / SPARC that are mainly deployed to run COBOL."

That's hardly surprising, but once again your lack of knowledge is not evidence of anything else.

Here at Micro Focus (founded a year before Oracle), our main products are COBOL implementations for x86 and RISC platforms. Sure, that's only a few hundred million dollars a year, which is no doubt small change to you financial-industry folks (many of whom are our customers). But it represents a goodly number of systems.

You're welcome to check out the web site and learn something. There are over 100 customer success stories currently published there.

2
0

Misleading!

Oracle omitted some hugely expensive parts of their benchmark configuration from the cost comparison. You can see the details at...

http://smarterquestions.org/2011/10/challenging-oracles-sparc-supercluster-claims/

2
0
Anonymous Coward

Oi, HP & IBM!

Postgres Oracle shim. You know you want to.

Opensource is a classic way of devaluing standard software.

Also, could someone enlighten me about the obsession with massive single-image database hosts?

Disk is cheap, 8-core hosts/blades are cheap. There must be an architecture somewhere which allows simple replication/sharding.

2
0
Boffin

@AC 06:33 GMT

Whatcha talkin bout Willis, like IBM mainframe storage groups????

0
0
FAIL

Will History Repeat Itself?

Many years ago Larry Ellison promised to show Microsoft up for the "purveyor of second rate software that it really is" (or words to that effect).

So having acquired the tools to do so with (OpenOffice, MySQL) he proceeded to destroy his side of things!

Even money says he'll now do the same with hardware.

4
0
Silver badge
Boffin

RE: Oi, HP & IBM!

"......could someone enlighten me about the obsession with massive single-image database hosts?...." It's about performance and simplicity of operation. With a large SMP server (or usually a cluster), all the interconnects are internal and so management of the server is management of the interconnects. With large clusters of smaller servers working in parallel (like Beowulf, as an example), there are lots of servers to manage as individual systems, and then the added work of the interconencts (usually high-speed NICs like Infiniband, 10GbE or exotic stuff like Hyperfabric, and the associated switches or hubs). As the parallel cluster grows, the management load grows by a factor two. I did a project a few years back where we did some benchmarks using a large Beowulf implementation against an SMP cluster, and although we could get almost as much application performance on the Beowulf setup for a lot less cash, the problems of keeping it up and running were a massive headache compared to the SMP cluster. With virtualisation now giving most of the large SMP vendors the ability to make their big servers look like lots of little ones if required, but still maintain that management advantage, the SMP option is still very attractive.

2
2
Silver badge

Foreigner Nuke Design?

I've worked in many places where, barring end of month and end of year where massive processing runs (that could have been trickled through easily enough but for 'adjustments') actually used the expensive boxes to maybe 5% capacity for 95% of the year.

Even then these processing runs could have been shortened considerably by not blindly adhering to relational database structures and therefore recalculating the total for every transaction on every run!

Ellison and IBM seem to make their money (like Colemans) from what’s left on the plate - companies 30-40 times their IT needs so the accountants can fiddle, sorry adjust, things.

0
0
Paris Hilton

When he is done

I remember watching in YouTube a vid where someone put together all the times Larry said 'Red Hat' in a conference.

Kind of funny, he was going to sink Red Hat, and yet Red Hat keeps reporting record quarters just about every quarter.

Another memo that got lost I guess.

Paris as she also seems to get a lot media attention for no real reason.

3
0
Big Brother

SInk Red Hat

Sink Microsoft

Sink IBM

LPOD will have the ministry of truth on fix his past pronouncements when he wins the war with Oceania

1
0
Joke

Larry Ellison is (thinks he is) Tony Stark...

Forget POWER-7 and T-4, what about those Ironman suits...

;-)

2
0

He needs all the money he can get

For his swanky new volcano layer, cat and private army.

2
0
This topic is closed for new posts.

Biting the hand that feeds IT © 1998–2018