MW-E2ED Call 3-May-2006

*Attendees*

Mark Poepping, CMU (stand-in chair, scribe)
Walter Wong, CMU
Kevin Miller, Duke

*Discussion*

. Mark, Walter and Kevin talked through the slides and datat model that Walter proposed in the slides - some questions but mostly seemed a good model to start with experimentation.

Next steps:

. Maya and Walter will continue work, this time Maya will begin working toward producing some records that will help to map to the model presented. As yet it's not clear whether those will result in simple syslog dumps or sophisticated state-based processing - most likely something in between, the notion is to start an iterative process of refining the record definitions to map to maximum usefulness with reasonable flexibility while managing overall complexity (if work needs to be done', maybe start by doing it where it seems easiest) - mom-and-apple-pie, but we have to start somewhere, and we do so knowing full well that we are certain to change it.

. The iteration on record breakdown and composition by necessity will require an interaction with analytics that have not yet been written nor tested for relevance (though many have been conceived). The first of these analytics were (as I recall) all based on 'reporting' scenarios. As such, some storage agent (or mock-up thereof) will be necessary before we can do 'real' evaluation of the design of the CER's. As I recall, we talked about a 'simple' mySQL db-loader that might be used to bridge us to support for experimentation, but I couldn't recall where we were on the storage agent efforts so... this is a request to Chas for more information on the storage agent plan and a suggestion for the next call to finalize how we'll bridge from the Walter-mail-CER's that Maya will drop onto the backplane, to the situation where we can experiment with analytics (to complete the loop for iteration on CER design). These experimental needs may be easily satisfied, we just need to decide how we'll do it (and who and when). This effort will bump into that quickly.