Py Bay 2016 - a First Report

PyBay held their first local Python conference this last weekend (Friday, August 19 through Sunday, August 21). What a great event! I just wanted to get down some first impressions - I hope to do more after the slides and videos are up.

First, the venue and arrangements were spot on. Check the twitter traffic for #PyBay2016 and @Py_Bay and you see numerous comments confirming that. And, I must say the food was worthy of San Francisco - very tasty. And healthy. With the weather cooperating to be nicely sunny around noon, the outdoor seating was appreciated by all who came from far away. The organizers even made it a truly Bay Area experience by arranging for both Yoga and Improv breakouts. The people were great - volunteers, organizers, speakers, and attendees. Props to all.

The technical program was well organized, and I’m really looking forward to the videos for the talks I couldn’t attend. Here’s some quick highlights that I hope to backfill.

  • OpenTracing - a talk by Ben Sigelman - big one for distributed systems, it promises a straightforward way to identify critical path issues across a micro-service distributed architecture. Embraced by a number of big companies (Google, Uber), it builds on real world experience with distributed systems.

    Programs just need to add about 5 lines of setup, and one call per ‘traceable action’ (whatever that means for your environment). The output can be directed anywhere - one of the specialized UI’s or traditional log aggregation services.

    There are open source, commercial offerings, and libraries for many languages (Python, Go, etc.) & frameworks (SQLAlchemy, Flask, etc.). As an entry level, you can insert the trace calls and render to existing logging. The framework adds guids to simplify tracking across multiple hosts & processes.

  • Semantic logging - a lightning talk by Mahmoud Hashemi was a brief introduction to the lithoxyl package. The readme contains the selling point. (Especially since I took Raymond Hettinger’s intermediate class on Python Friday, and he convincingly advocated for the ratio of business logic lines to “admin lines” as a metric of good code.)

  • Mahmoud Hashemi also did a full talk on profiling Python performance in enterprise applications, and ways to improve that performance. (And, yes, we write enterprise applications.)

And there was lots more that I’ll try to cover later. And add in some links for the above as they become available.