What's going on: HP Technology Forum and Expo
By Jean-Pierre Dijcks-Oracle on Jun 15, 2009
Should you be in the neighborhood, come visit Hamid and me at the HP Technology Expo in Las Vegas this week. We'll be sharing details on HP Oracle Database Machine with the folks out there in a set of workshops.
We'll be going through the database machine and explain how things work - in detail. One part of the show is using an interesting demonstration framework with which we can show the value proposition of Exadata.
Here is a sneak preview of some of those demonstrations.
We are focusing on three aspects of Exadata - apart from the completely configured out of the box benefit:
1) Linearly improve performance by adding hardware
2) Linear scalability to grow with your enterprise
3) The smart scan advantage
In the first scenario we are doubling the hardware (in this case we are doubling the number of Exadata storage servers) while keeping the data volume constant. The screen here is showing a nice graph on the execution times on the in-house system:
With the same data volume, the same query runs twice as fast on double the hardware.
In the second scenario we are doing something similar as happens in your data warehouse on a weekly or monthly basis. We are doubling the data. To show Exadata scales linear, something you want in your data warehouse, we add double the hardware as well. Then run the same queries.
The result is better than linear scalability in most cases. On the in-house system, you see the example with random allocation on disk. In other words, we did not make the larger system space go faster by positioning data only on the faster outer rim of the disk.
For a last sneak peek, we are running the database machine with and without smart scans enabled. The result of using smart scans is that we move a lot less data over the internal network to the database servers. Smart scans do all their work on the storage servers and allow off-loading of where clauses (and lots of other stuff) to those machines. Consequently the data returned is only that set of rows and columns (!) that the database needs to aggregate the results.
As you can see in the graph, the volume of data moved is significantly smaller than the volume of data scanned.
The combination of these three and the addition of an out-of-the-box experience is what makes Exadata and Database Machine such a nice product for the warehouse community. Once in Vegas I'll try to keep you posted on what is going on. It should be quite an exciting show!