Real Application Testing - RATs!!

I attended Oracle Day San Francisco today. There were some really fine presentations. One was "Managing Change with Oracle Database 11g" which mainly talked about "Real Application Testing". My opinion was that it seems like a good idea, but that it isn't mature yet.

In Real Application Testing you capture the actual production workload for replay and comparison on a test system. Nice idea, but the proposed test system is the same size, and has a much machinery and disk space behind it as the production system. I don't know about you, but I have never seen such a test system as big as the production system.

The lecturer suggested that

• You could discount any disparities that are to be blamed on the platform difference.
• You could run the workload on the test system against the old application, and then run the workload against the new application (or database version) on the test system and make a non production apples to apples.

Either solution is a good idea, but not perfect.

There was a lot of emphasis given that a production workload is much faster to prepare than a test workload,(created by testers) complete with charts to prove the point. But a production workload is going to be full of social security numbers and credit card numbers, etc. The presenter suggested that you could mask sensitive data, but there was no component for this item in the time line that compared traditional testing with Real Application Testing. I have just read in Lutz Hartmann's blog that Oracle was announcing in November that the ability to mask 'was coming'. I don't know if it is here yet.

Overall, Real Application Testing seemed like a good idea that will be much better by Oracle 12 or 13.