This month's most interesting story from the Windows NT Magazine Lab won't appear in the magazine. You should be reading about the Lab's tests of Oracle8. But before I tell you why you're not, I'll update you on developments from past columns.
A Benchmark by Any Other Name ...
If you were unhappy about the Lab's benchmark report last November that compared NetWare with Windows NT (Carlos Bernal, "NT vs. NetWare: File Services Grand Prix")--you'll be pleased to learn that Novell has agreed to work with the Lab on a follow-up test. Working with Novell will eliminate any implication of bias and concerns that the Lab's hardware or software test configurations are unfair.
The Speed I Need
Regarding my quest for ISDN connectivity, US West finally installed the ISDN line in my small office/home office (SOHO) lab. (I had waited just a few days shy of 6 months to receive the line.) The 3Com/U.S. Robotics software I used to configure my I-modem worked without a hitch. Maybe the Fates smiled on my I-modem configuration, knowing I had waited so long to get the line installed.
The Zen of Oracle
Now, on to this month's topic: a report on Oracle's very large memory (VLM) implementation for Oracle8 on Alpha systems. Whereas NT Server 4.0, Enterprise Edition (NTS/E) limits VLM support to 4GB of memory (only 3GB of which is available to the VLM application), Oracle has created a VLM implementation that supports 8GB of memory. At conference after conference, Oracle has demonstrated how Oracle8 technology uses VLM to speed up processing by caching information in the VLM area instead of on disk. These tests show that processing is 15 times faster when VLM is turned on.
I'm not saying the Lab Guys don't believe the test results. We simply wanted to verify these results in a controlled Lab environment.
A major Alpha server manufacturer volunteered to help and sent us a 4-way Alpha system with Oracle8 installed and 8GB of memory. As happens in all large benchmarking projects, we ran into some snags. So, we turned to our hardware vendor and Oracle for help. At first, Oracle and the hardware vendor were cooperative. However, one day our main Oracle contact stopped talking to us without an explanation. He said that if we needed more information, we'd have to talk to Oracle's marketing department.
While we scratched our heads over that mystery, a representative of the hardware vendor called to say that the vendor would have to withdraw from the test and we needed to send the server back by the end of the week. What happened? After doing some digging, I learned that an Oracle executive discovered our project and killed it because Oracle doesn't officially encourage, support, or endorse benchmarks that publications conduct. I had heard rumors about Oracle's no-benchmarking policy, but I was still shocked to face it. I asked my Oracle marketing contact about the company's policy, and the reply was, "Because we have nothing to gain. We are already the market leader, so what's the point?"
In other words, I'm supposed to believe that Oracle8's performance is superior to SQL Server's performance (and the performance of other databases) based on the absence of analytical data. I know Oracle's chief executive officer is fond of Eastern cultures, but this policy is a little too Zen-like for me. A Zen master might be able to judge the depth of a pond by its stillness, but I can't judge the performance of a database by its features.
I'm no master, but I do study a little Zen now and then. I have a Zen saying posted in my office that I'd like to share with Oracle: "No more solutions! No more words!" This saying means that when you weed out solutions and words, you are left with only action. Actions speak louder than marketing brochures. Oracle needs to stop stonewalling and let us prove that Oracle8 lives up to its reputation.
