My first presentation on zipkin took place at Betahaus in Barcelona, hosted by Schibsted at the Netflix OSS Meetup.
This deck overviews latency anaysis, and how tracers substitute for people gathering and organizing things from server logs. The resulting tracing systems give you an upper hand on determining root cause even as calls cascade across a network.
Being a first run, I learned some things about the deck, which I'll change in later versions.
Firstly.. less slides. This is too many slides for 50m, so some hard choices need to be made.
It might be better to move talk about tracers until after demo'ing zipkin. Not everyone are interested in tracers: multiple people asked if there was a prana side-car or similar that they could apply to an existing app to make it instrumented.
I think that this could be better organized by demoing zipkin, showing an implicit tool like sleuth.. *then* discussing how that works (ex all the cool bits about tracers and how they in-band vs out-of-band). I was missing a diagram on some of this, which I literally drew on paper there :)
Another thing the group noticed was we could touch on automated analysis. Even if a single slide, mention what insight people are thinking about here, ex. annotating spans that are notable.
Unsurprisingly, folks were unimpressed with zipkin's UI, even if feedback was kind :P This is known, but we don't have anyone who's offered to help. At least we know it is still an issue!
Finally, and more important than the notes above, was that the crowd were quite patient and giving with feedback and ideas. A first deck is like telling jokes at amateur night (I think.. as I've not tried it). The Netflix OSS crowd in BCN rolled with my mistakes, gaps in slides and mental glitches. They were quite astute throughout and engaged with feedback. I'd love to come back and present here again!