It is likely that most readers of the bayon blog are also readers of rittman.net. However, just in case someone missed his posts on tuning OWB Performance Tuning Framework I highly recommend that you read/bookmark them. For the casual OWB user this might not be necessary; developers who deal either with a great many mappings or mappings that process significant data volumes this is a “tool” to keep in your toolbox.

Mark is on to a great method; the application of battle tested tuning techniques to an often used but rarely fully understood tool. I’m not sure how much time I’ll have over the next few weeks, but if I do perhaps I’ll throw my oar in as well. Things I can think of to contribute to the community dovetailing on Mark’s work include:

  • Unify the approach of monitoring the Clock Time and drill down to the performance data Mark has retrieved. ie, build a set of reports (or even a small data mart!) that consumes the public runtime performance views in the OWB runtime repository. I’m a big fan of quantifying the benefit of a tuning excercise. Perhaps a way to track which processes/mappings are slowing your loads, elapsed/time versus volume (ETL throughput), etc. Consume and understand the basic metrics, then drill down on a particular mapping into the information that Mark has developed.
  • An OMBPlus implementation of the “on/off” trace flag Mark refers to :The way of enabling tracing should either be built in by default, or easily added by the OWB developer. Ideally it should be simple to switch it on or off (perhaps levels of event 10046 tracing?). I bet one can create a script that adds/removes the pre/most mapping process that Mark has added manually to the mapping. In this method we could programmatically turn tracing on for a defined list (from one to entire repository) of OWB mappings. The logic to set the key for the process flow operators using OMBPlus might be a wee bit more difficult (would have to iterate through operators in a process and possibly make changes). The biggest drawback, and I don’t see a clear solution at this point, is how to change tracing at RUNTIME instead of deployment. Making the changes via OMB requires a redeployment but it would be most beneficial if tracing was in the code, and changed at runtime.

A great start of the conversation that’s been long overdue… Perhaps a small OWB coming of age? The technology is now used widely enough to warrant some of it’s own methods, patterns, and strategies for not just building but rather building well.