Yes, I feel very fortunate to be able to work on that system. There are stats for seeing some of the unique features such as offloading (how much data is actually transferred, etc..) and what’s returned from flash cache vs. actual disk access. Although the stats are not quite as easily accessible as we are used to for non-Exadata stuff (i.e. not all of it is accessible at the statement/cursor level via V$SQL). But there are new wait events, new plan options (which are available via xplan, etc…) I intend to share that info as soon as I have some time to put it together. On a side note, we have our own Exadata showing up in a couple of weeks. So we should be able to do whatever kind of testing we want, without having to worry about “real” work being done on the platform as well. That should be fun! Although there is no substitute for being able to look at a real production system, so as I said earlier, I am really thankful that we have had the opportunity to work on the real system (which is in limited production now by the way).

Kerry

This is some stuff I have been itching to put my hands on.

It might be great if you can test both DW queries ( that are great for offloading) and some ETL processes performance. I would be delighted to see cell offloading in work, execution plan differences, statistics differences etc.

In the non-exadata configurations, we use consistent gets, db block gets, db time etc to resolve performance issues. How is that going to change with exadata? how effective are the tools that we currently use, in exdata environment?

Lots of questions, so little time !

Kevin dude,
you are humble and you had lot more than 6 attendees in your hotsos presentation. Good stuff, that you present.

Cheers
Riyaj

The main advantages of Exadata are #1) database performance and #2) ease of deployment. The Oracle Database Machine comes from the factory as a pre-built system of servers, storage and networking. Oracle installs & configures the machine, and the system goes from the loading dock to loading data in about 2-3 days. The system performs better than traditional hardware architectures because the Oracle software is tightly integrated into the hardware. Part of the RDBMS code actually runs INSIDE the storage tier. The software inside the storage knows how to run fragments of SQL statements, and it knows how to utilize the built-in cache to best advantage for database performance. Oracle software doesn’t run inside of a traditional disk array, so you simply cannot get this degree of integration without Exadata.
-Hope this helps…
Chris

Will be more usefull if you could provide some insights with test cases on advantage of using Exadata V2 compared to absence of this technology.

Also some basic stuff with examples will more helpfull as some DBA’s are not so lucky to work on this technology.

-Yasser

Flash cache in the Database Machine hosts would be of no value… trust me… I know… in fact, if you had joined those other 6 attendees in my Hotsos Symposium 2010 session you’d know why :-) Hold it, you can get the slides since you were a speaker…

I plan to do a contrast presentation webcast-style on the topic of these two different Oracle Flash technology offerings. There are a lot of DIY types that seem to suppose that flash is flash so they can back the “upstairs” with flash and dangle regular old spinning rust for the downstairs and voila the product is a self-made Database Machine. The problem is, it isn’t so.

That avatar is a photo of a friend’s Jeep circa ’49 or so. It never moves. We hang our deer in the same building though when successful on a hunt if that’s any consolation :-)

Yes – 2 quarter racks. I was thinking Flash Cache in the DB Server might be interesting as well. Yes, screaming fast on the big tables!

What’s with the picture by the way? My eyes aren’t that great anymore but it looks like a jeep with a deer blind built on the back.

Kerry

…you must be working with a customer that has either a half or a quarter-rack. Thoughts of larger form factor hosts would seem feasible in that case I suppose. A full-rack configuration has no spare room to “fatten” the hosts.

…glad to see you’re getting to have some fun. Fast stuff, eh ?