Ellison rips IBM, shows off Sun-Oracle benchmarks, offers $10M prize

Ellison rips IBM, shows off Sun-Oracle benchmarks, offers $10M prize

Summary: On Sunday night Larry Ellison took the stage at his annual event, Oracle OpenWorld in San Francisco, and came out swinging like a samurai warrior with a newly-sharpened sword. His main target: IBM.

SHARE:
TOPICS: IBM, Hardware, Oracle, Servers
23

On Sunday night Larry Ellison took the stage at his annual event, Oracle OpenWorld in San Francisco, and came out swinging like a samurai warrior with a newly-sharpened sword.  His main target: IBM.

Ellison (right) told a capacity crowd of thousands of IT professionals that IBM has been running a campaign called "Sunset" in which it has been telling Sun customers that Oracle is going to get out of the hardware business and therefore customers should start moving their software over to IBM servers.

Ellison attacked this idea. "We are not selling the hardware business," he said. "No part of the hardware business are we selling."

Then he actually displayed some of IBM's own ads - a bold and highly unconventional move since it gives IBM free attention - before ripping them to shreds. Without going into the gory details, Ellison basically said IBM deploys more servers than necessary, that IBM servers are power hogs (even though IBM talks a big game about being energy-friendly), and that a Sun-Oracle combination runs circles around IBM in performance.

Ellison showed off a new benchmark that reported:

  • Sun-Oracle server configuration used 8x less hardware than IBM (in this comparison)
  • Sun-Oracle server config ran 26% faster than IBM
  • Sun-Oracle server config consumed 4x less energy than IBM
  • Sun-Oracle server config had 16x better response time than IBM (1.22 seconds for IBM vs. 0.08 for Sun-Oracle)

I'm always wary of these types of benchmarks because they usually involve one company's systems being highly-tuned and then comparing them to a competitor's standard configuration. Nevertheless, Ellison said that he is so confident that this comparison will hold up that he's launching a new program in which Oracle will pay customers $10 million if a Sun-Oracle configuration isn't at least twice as fast as a comparable IBM solution.

Ellison also gave a parting shot to IBM's eco-friendly "smarter world" campaign. He said, "I don't know what building a smarter world means. We're going to focus on building smarter computers."

He also implied that IBM was being hypocritical since its servers looked like energy hogs in the recent benchmark. "Sun's processor is called SPARC... IBM's processor is called Power. Now we know why," quipped Ellison.

The combative rhetoric against IBM shows that Ellison is seriously worried about losing Sun customers while the Oracle-Sun deal awaits regulatory approval. In September Ellison told the Churchill Club that Sun is losing $100 million a month while they wait.

Here are some other notable Ellison quotes from the keynote:

  • "There are limits to how far you can go if you just do software."
  • "Whatever you think of Apple, I think they've done a great job of tackling the hardware problem while tackling the software problem."
  • The goal of a combined Oracle-Sun will be to "integrate hardware and software and deliver revolutionary systems."
  • "Solaris is the number one enterprise operating system in the world."
  • "We're going to spend more, not less, on MySQL... We're going to increase our contribution."

Topics: IBM, Hardware, Oracle, Servers

Kick off your day with ZDNet's daily email newsletter. It's the freshest tech news and opinion, served hot. Get it.

Talkback

23 comments
Log in or register to join the discussion
  • RE: Ellison rips IBM, shows off Sun-Oracle benchmarks, offers $10M prize

    Fairly typical of desperation attempts, use words
    not real proof through history.
    wadej915@...
  • "Unbreakable"

    And that one did not turn out too well...
    GuidingLight
  • RE: Ellison rips IBM, shows off Sun-Oracle benchmarks, offers $10M prize

    I know of at least 1 company running their corporate DB2 server on SPARC Servers. About 5 years ago there used to be an IBM "tiger team" that would go around telling customers that DB2 would run better on SPARC/Solaris - I believe they were all reassigned or RIF'ed. I heard that it has been suggested that this company migrate from SPARC to a mainframe!?
    jameswlucas
  • RE: Ellison rips IBM, shows off Sun-Oracle benchmarks, offers $10M prize

    In the grand scheme of things, Ellison ripping ANYONE is not
    particularly newsworthy. He ripped on God, for pete's sake.
    When I worked for Oracle, the question was always, "who is
    'Uncle Larry' setting his sights on this year?" But enough of the
    persona, what about the product?

    IMO, Oracle database still is a solid product. Sun servers and
    Solaris OS are too. A well-tuned Oracle database running on
    Sun HW/SW platforms beats IBM hands-down. But, that was
    while Sun was an independent entity. Now that Larry is poking
    his fingers into the hardware space, the question stands, "why
    do you want to be an IBM wannabe?" That vision violates the
    Marketing Prime Directive -- If you can't be number one where
    you stand today, carve your own niche and be number one
    there.

    Larry,
    IBM is already Numero Uno in the combined HW/SW/Support
    Services space. Better break out your carving knife.

    mjstelly
    • Oh really ?

      "A well-tuned Oracle database running on Sun HW/SW platforms beats IBM hands-down."

      I must admit I have had the total opposite experience, at least when we are talking about machines in this millennium.

      // Jesper
      JesperFrimann
  • Hey Ellison.. Read this

    I have just now gotten my feet wet with Oracle. Let me tell you this. There is not one shred of Common sense built into the Oracle Database System. As a developer, it takes twice as long to do something, or 3x as much cost wise to get a professional to do it for you. Oh, and good luck finding documentation that you can use.

    Go read their forums, their tech support forums, or their wiki.. It's crap. Read thru some of the responses customers get.. more crap. Honestly, Larry needs to put some money into supporting his software and growing his brand that way. DB2 and MSSQL are tons better than the Oracle Software for a reason.. They have the best documentation out there for their product, and there are enough users available to share their knowlege.

    So.. um.. Hey Ellison: Quit wasting your breath fending off attacks.. and start using it to tell your people to document their software.

    DocNasty
    • umm...not so fast!

      If online help becomes too good, then how would Oracle justify those inflated prices for paid 'support'?
      Linux Geek
    • Are you serious ....

      You call yourself a developer but you don't want to *learn* how to use proper database product!

      Maybe you should stick to Access (or SQL Server aka Acess++) and leave the real systems to people who could be bothered to learn, use and implement enterprise database systems.

      PS Any Oracle developer worth their socks will be able to direct you to docs.oracle.com but I guess that's also something you learn.
      ranchHunter
      • You must be kidding me!

        I've used Oracle for nearly 8 years and I beg to differ. Oracle has a huge knowledgebase and repository of information, not including dozens of websites dedicated to Oracle technology in general.

        If you are unable to find something specific you should ask instead of making broad assumptions about Oracle. Have you looked at asktom.oracle.com in addition to what "ranchHunter" pointed out?

        DB2/MySQL are all good products and this is no way a reflection on them but the Oracle Database product can easily hold it's on against any database in the market.

        Do some research and be willing to invest some significant time undertanding the technology before you judge the software or the company.

        Thanks.
        thewhiteknight
  • RE: Ellison rips IBM, shows off Sun-Oracle benchmarks, offers $10M prize

    I don't suppose that there is a technical organization somewhere that could put together a some comparative test?
    Maybe that could show everyone information about the pros and cons of each companies product and the strengths and otherwise of different configuration.
    Informed decision making could become a whole new trend.
    Agnostic_OS
  • oracle/db2 on sun/ibm

    I've been an Oracle DBA for over 10 years and a DB2 DBA for about 8 years. I've run Oracle on Sun hardware and IBM hardware. I've run DB2 on Sun hardware and IBM Power-series hardware.

    I've got to say that DB2 on either platform, Sun or IBM, has been rock-solid. I wish I could say the same for Oracle. DB2 databases are easier to administer properly than the Oracle databases (IMHO). IBM support beats Oracle's hands down.

    Ten years ago I never would have caught myself saying that, I was too big of an Oracle fanboy.
    mark000
  • RE: Ellison rips IBM, shows off Sun-Oracle benchmarks, offers $10M prize

    Wow, Ellison looking all fake-baked & what not. He should be doing Just For Men commercials with Keith Hernandez and Walt Frazier. I still have memories of Larry Ellison personally suckering Marconi Communications into a miserable failure of an Oracle E Business Suite implementation back in 2000/2001. How is he going to keep both the acquired Sun Micro and HP happy?
    parkerjgpatton
  • RE: Ellison rips IBM, shows off Sun-Oracle benchmarks, offers $10M prize

    Looks like Ellison lined up to be second in line behind Microsoft and IBM staying stumph, no doubt treading lightly and carrying a big stick.
    SPARC servers and Solaris were good earners for me as opposed to installing a mainframe - starting with delivery at 07:00, an hour for lunch in the pub and handing it over to the customer by 16:00, I/O attached and ready for the IPL. It would take days to get a large SPARC system to the same point.
    sboyce@...
  • RE: Ellison rips IBM, shows off Sun-Oracle benchmarks, offers $10M prize

    Oracle's PR machine dropped the ball. Oracle should have
    been ahead of the game.
    david@...
  • RE: Ellison rips IBM, shows off Sun-Oracle benchmarks, offers $10M prize

    Well, the worst part is the price of the config. I mean they don't even price in what it costs to buy the software.
    Basically it is what it cost to have/lease the Database
    software for a period of 3 years.

    And also they aren't using ordinary Oracle support, the 3 year of support is Metalink only. I mean.. honestly.

    If they had done it properly then even with a think discount they would have gone way past 3$/tpmc.

    And try to calculate the single threaded performance on that cluster, I mean it's 2512 TPMC/ thread. That is Pentium Pro 200Mhz throughput from the last millennium kind of throughput.

    Bleh

    // Jesper
    JesperFrimann
    • That makes no sense...

      [i]"And try to calculate the single threaded performance on that cluster, I mean it's 2512 TPMC/ thread. That is Pentium Pro 200Mhz throughput from the last millennium kind of throughput."[/i]

      That argument makes no sense. In a nutshell, you are saying that since a bus carries passengers from point A to point B significantly slower than a car, everyone should stop taking the bus and drive their own car.
      The whole point of multithreading is to increase performance while reducing power consumption by spreading the load across as many threads as possible. Your argument is basically "Cars are faster so everyone should drive cars" but ignores the fact that the bus uses less fuel per passenger than any car for the same trip.

      Of course that's just my opinion, I could be wrong.
      914four
  • RE: Ellison rips IBM, shows off Sun-Oracle benchmarks, offers $10M prize

    "That argument makes no sense. In a nutshell, you are saying that since a bus carries passengers from point A to point B significantly slower than a car, everyone should stop taking the bus and drive their own car."

    It does make sense. I think you have gotten it the wrong way around. The T5440's used in this benchmark uses 3072 threads that are each capable of doing 2.512 transactions per minute. Or it uses 3072 busses that each can transport 2512 passengers per time unit.
    If you took for example number 2 on the tpcc list, a power 595 then it uses 128 Busses/threads that each can transport 47.540 passengers per time unit.

    "The whole point of multithreading is to increase performance while reducing power consumption by spreading the load across as many threads as possible. Your argument is basically "Cars are faster so everyone should drive cars" but ignores the fact that the bus uses less fuel per passenger than any car for the same trip."
    Nahh.. it is you who are claiming that 3074 slow cars are better than 128 faster busses.

    "Of course that's just my opinion, I could be wrong."

    There is nothing wrong with running things in parallel.
    But for a really good useful system you want to have both, both good single threaded performance and the option to spread your load over a good chunk of threads. And you also want fairly stable response time.

    If you compare the top TPMC performers of the different processor architectures you can see that there is a huge difference between T2+ and the others.

    T2+ 7,717,511 tpmc 2512 per thread.
    power 6.085.166 tpmc 47540 per thread
    x86 1.200.632 tpmc 25013 per thread
    IA64 4.092.799 tpmc 15988 per thread.

    // Jesper
    JesperFrimann
    • I don't get your argument...

      Where do you come up with the notion of 3072 buses? This is one server (bus) carrying 3072 passengers vs another bus able to carry only 128.
      Again, your argument makes no sense.

      I just finished listening to the McNealy/Ellison Oracle World Keynote speech and one thing the article does [i]not[/i] mention is that the IBM solution took up [b]76 standard 19" racks[/b], used 6 times the electrical power of the Oracle/Sun solution based on number of transactions and was [i][b]not fault tolerant![/i][/b].

      The Oracle/Sun solution used 9 standard 19" racks, offered 25% better throughput and [i]16x faster response time![/i]. And it had no single point of failure. Sorry, but I don't buy your argument. Even if Larry was exaggerating, 9 racks vs 76 is not something you can easily fudge.
      914four
      • Different technology generation

        The new benchmark uses 1.92TB of flash drives and 8Gbs disk controllers. These were not economical when IBM published the number 2 TPC benchmarks over a year ago.

        It will be interesting to see an IBM benchmark with similar storage gear using the current DB2 V9.7.

        DBMS comparisons can only be made if both are running on the same underlying hardware. That happened in 2005 when DB2 and Oracle published benchmarks on the same P5 hardware (number 5 and 9 of the TPC top 10 performance). At that time, DB2 came out 2X faster.

        With Larry's challenge is he waiving the Oracle license provision that prohibits customers from publishing benchmark results?
        Norm_z
        • Thanks Norm

          Jesper was saying that the T2+ CPU was slower than the Power 6 in the 595 based on how quickly it could process a single thread, and my argument was that it shouldn't matter, it's the overall result that counts. The test of DB2 vs Oracle proves that DB2 runs faster on Power, but since IBM doesn't need to tune DB2 for HP-UX, Solaris, Tru64, Red Hat, Suse, Windows, etc (DB2 on Linux and AIX are different databases, whereas Oracle uses the same coding ported to each OS), it's not really a fair comparison. If you want to do a drag race, then DB2 may be faster, but a real world test is using the database to perform a common task using the hardware it is best suited for. Just because a top fuel dragster can do a quarter mile in 5 seconds does not mean it is better than or even comparable to a Porsche Turbo. I'm thinking wait and see, but even with a year's difference in technology we are still talking 1/8 the footprint for more than double the performance, which blows Moore's law out of the water.


          [i]"With Larry's challenge is he waiving the Oracle license provision that prohibits customers from publishing benchmark results?"[/i]

          I'd never heard that; I know EMC prevents customers from publishing benchmarks or even publicly acknowledging any failures (they threatened to sue one of my clients' New York City office manager) but I've never heard of Oracle doing anything like that. My experience has been that they encourage "bake offs" where they can show their performance advantage, at least with HP-UX. Of course they have been known to fly in the tiger team to tune for the win (your personal mileage may vary) but I've never seen them say anything about not publishing benchmark results. They may prohibit publishing the benchmark [i]without first contacting them to describe the test parameters and results to make sure the tests were done correctly [/i], but that's no different than what many others including Microsoft, Borland, CA or VMWare do.

          It will be interesting to see where this goes, but after watching the keynote, I do wonder what IBM's eco story is.
          914four