Tuesday, February 02, 2010

Larry Sure Knows How to Get Press

Going under the assumption (I assume) that no press is bad press, Oracle CEO Larry Ellison has attacked IBM's DB2... but he made several factual errors in his rant.

Here are some of the highlight (?) of the claims Ellison made about DB2 during a webcast last week.

Regarding TPC-C benchmarks, Ellison claims to have "(blown) the doors off of IBM. We crushed them." He went on to elaborate saying "In a machine that took up less than 10% the floor space, of IBM's record setting computer. We ran faster, we ran a lot faster: using a tiny fraction of the floor space, a tiny fraction of the power, cost less."

First of all, technicians working in trenches know that benchmarks are not indicative of real life performance. That aside, it is true that Oracle currently has the leading TPC-C benchmark result. Until late in 2009, DB2 enjoyed a massive 49% lead over Oracle. Oracle's most recent results give them a 25% lead (using more than six times as many CPU cores to do it).

Regarding the claim of using less space and power, this is due to Oracle using flash memory and comparing it with an IBM benchmark using conventional disk technology. If Oracle compared its benchmark to an IBM system using flash memory, these claims would not stand.

Later, Ellison claimed that "SAP chooses the Oracle Database to run under SAP in almost all their large accounts." As anyone who follows the computer industry knows, this claim is rather absurd. SAP's customers choose the DBMS to run, not SAP. And if SAP had anything to say about it, they would not recommend Oracle, their biggest competitor in the commercial business applications space. Furthermore, SAP favors DB2 for their own systems. They operate more than a thousand SAP systems, and all of those systems run on DB2.

Perhaps the silliest of Ellison's comments is this: "The Oracle Database scales out, IBM DB2 for Unix does not. Let me see, how many servers can IBM put together for an OLTP application? Let's see, how many can they group together? Um, one. They can have up to one server attacking really big jobs. When they need more capacity, they make that server bigger. And then they take the old server out, put a bigger one in. And when you've got the biggest server, that's it. That's all the can do for OLTP." Ellison also claimed that IBM "can't scale out, they can't do cloud, they can't do clusters, the can't do any of this."

I bet this surprised a lot of DB2 users doing these things with DB2! DB2 Parallel Edition was released in 1995, along with the capability to scale to a system of over a 100 Unix servers. DB2 LUW scalability is proven in many of the world's largest OLTP environments. Consider this press release talking about how DB2 LUW powers one of the largest OLTP systems in the world.

And what about that clustering claim? Evidently Mr. Ellison slept through 2009. IBM DB2 pureScale, released last year, offers powerful, efficient database clustering. For a cluster of 64 nodes, DB2 pureScale maintains 95% efficiency. At 128 nodes, DB2 pureScale maintains 84% efficiency. This is important because if you are growing a cluster to handle bigger workloads, you want your hardware to be doing productive work, not handling system overhead. On the other hand, Oracle RAC has a 100 server limit...

Ellison also made other far-out claims about IBM like "They're so far behind, I don't think they have any chance at all. I'm serious." Ellison also said "They are not competitive in the database business, except on the mainframe."

If this were true, why would Ellison spend any time thinking or talking about IBM. He must be worried, IMHO. Anyone with even a cursory knowledge of the computer industry has to admire IBM. They have led the industry in developing patents for the last 17 years. In 2009, IBM produced 4914 patents while Oracle did not even place in the top 50 patent leaders. A search of the US Patent office database reveals 1588 patents with "database" in the patent description while Oracle produced only 184 patents.

Hyperbole is one thing, but gross inaccuracy is another. In his latest tirade, Ellison is guilty of both. Oracle makes a good DBMS... pity its CEO doesn't think it can sell it on its own merits without making up stuff about the competition.


Wednesday, October 28, 2009

IOD2009 Day Two

Day two of the IBM Information on Demand conference was just as informative and exciting as day one. The day kicked off with a general session titled "A New Kind of Intelligence for a Smarter Planet." The idea presented is that the world is changing. It is becoming more instrumented, interconnected, and intelligent. Basically, as Steve Mills of IBM clarified, the ability to embed intelligence into millions of things will lead the transformation to an information led smarter planet. And that this information-led transformation will create opportunities for organizations to strategically gain control of information and create a new kind of intelligence.

Expanded intelligence begins with sensors and metering, which is doable today because price points have become reasonable now. There are 1 billion transistors per human, today. And an estimated 2 billion people will soon be on the Internet. At the same time, we are moving toward one trillion connected objects (video cameras, GPS devices, healthcare instruments, and even livestock). And there are hundreds of satellites orbiting the Earth, generating terabytes of data each and every day.

As we begin to intelligently interconnect these devices and analyze their streaming data, a transformative opportunity can results...

IBM brought up three customers to talk about how their organization were helping to transform to a smarter planet. The customers came from Cardinal Health, Statoil Hydro ASA (Norway), and the Food and Drug Administration. Highlights of the panel discussion:
  1. Security in the supply chain for food is absolutely critical and food safety is one of the most complex systems to deal with.
  2. The US imports 50pct of its food - from over 150 different countries.
  3. Every food supplier to US (domestic or foreign) must be registered with the FDA.
  4. And each supplier must complete forms as to the safety of its food. This can be difficult since suppliers range from large agri-businesses to small farms many of whom do not have a computer at all. So some records are probably kept on paper in shoe boxes.
  5. Supply chain security is very important in the health care (drugs) and oil industries too! In fact, it was discussed how each of these seemingly disparate industries face many similar challenges.
  6. Preventing problems requires understanding risk. And complexity requires collaboration in order to succeed becase nobody has enough resources to do it all alone.
  7. In some areas (such as remote parts of Norway) instrumentation is essential to even get the data because nobody wants to go there.
  8. Legacy systems often are rich sources of information (hey, that probably means mainframes!)
  9. Analytics are required for prevention of problems, but also aid in reaction to problems that actually occur too. No one prevents 100 percent of their problems.
After the panel IBM came back and wrapped it up. They mentioned how IBM was awarded the National Medal of Honor for their Blue Gene supercomputer for DNA sequencing. It was a very informative and entertaining general session.

I then attended the DB2 9.7 versus Oracle 11g Smackdown presentation. It was chock full of statistics on why IBM's DB2 is superior to Oracle in terms of cost. The presenter explained how DB2
outperforms Oracle on TPC-C benchmarks for the same test, on the same machine, at the same point in time. He cautioned folks to to read the small print details on all benchmark results... for example, if you are examining the cost of ownership double check to see whether the benchmark uses a term or full purchase license. Also, the cost of the database license depends a great deal on your ability to negotiate a discount (if the vendor will discount). And you also need to be aware of how the products are licensed. Some features are separately licensed for both DB2 and Oracle. The bottom line is that licensing can cause a more than 30 percentt swing in price performance results

But do people even believe these benchmarks any more? I don't think very many people put much stock in benchmark tests today.

The author frequently cited an independent study by ITG that compares the value proposition of DB2 9.7 versus Oracle Database 11g. You can read the study yourself at this link.

(Note to my regular z/OS readers: the previous discussion was all about DB2 LUW and not DB2 for z/OS).

I also got to attend a special briefing for bloggers on IBM's new stream computing solution, which I blogged about on my Data Management Today blog if you are interested.

And finally, I'll plug my Twitter feed again. You can follow my tweets at IOD this week (well, thru Wednesday) by following me on Twitter at http://www.twitter.com/craigmullins.

Labels: , , , ,