People who have been around Oracle Exadata a lot since version 2 of the database appliance came out a couple of...
years ago usually have only glowing things to say about it. But at the first day of a two-day Exadata conference being hosted by Enkitec, they found some minor glitches.
First is the unpredictability of some performance traits, particularly the storage indexing. Kerry Osborne, a co-founder at the Dallas-based Oracle consultancy Enkitec, said, "They haven't built a lot of control into the storage indexing. It can be a little frustrating because you can try it one time and you get sub-second response time, and then the next time it might take 20 seconds."
Osborne added that most people in production won't see the variations, because once they get Exadata up and running, it's unlikely that they'll shut it down and flush out the storage cells that often. At Enkitec, where they have done a lot of Exadata testing, the practice is more common.
The lack of documentation made available by Oracle is another problem. Cary Millsap, founder of Southlake, Tex.-based Oracle software performance consultancy Method R Corp., said he thinks Oracle is making a "big mistake" by not offering more documentation.
More on Oracle Exadata
Check out our Exadata hardware guide
Can you really compare Exadata with SAP HANA?
No doubt there's a lot of interest in Exadata
"If you can't get to the document and can't get to the machine, then there's not much you can do," Millsap said. "I don't think Oracle is reaching the potential of the machine as much as they could."
Both Osborne and Millsap spoke during a conference on Monday called the Enkitec Extreme Exadata Expo, which continued through Tuesday.
Osborne and Millsap stressed that these were minor problems with a platform that they have found mostly amazing. Millsap said that from the moment Exadata was launched, even the Oracle-HP joint project that was the first version, he was impressed with how the engineers were trying to push the code base deeper into the machine and down to the storage level. He thought that was ingenious.
"The fatness of the pipe between the storage and the database buffer cache was impressive," he said.
Millsap also talked about his first impression of Exadata, saying that the hardware side wasn't what caught his eye. What he thought was exciting was the "Smart Scan" stuff, which, as he put it, would "push as much filtering as deep as you can."
Osborne said he has seen a lot of resistance to Exadata from what he called "long-term performance guys." He considers himself part of that group, a group that tends to think throwing a new hardware platform at a problem isn't a good solution. But Osborne said he has been convinced.
As powerful a machine as they think Exadata is, they both added that it is possible to max it out, just like any other machine.
"If you buy a $3 million box and put a little departmental app on it, it will absolutely scream," Millsap said. "But like any box, you're going to fill it with as much load as you can. There are ways you can bring anything to its knees."