Last week was the first time that I participated in the conference Berlin Buzzwords. Even though I read that the conference is all about store, scale and search I wasn’t quite sure what to expect, except for lots of information about Elasticsearch (the Gold partner of the conference).
After having attended the conference I can say that participating at Berlin Buzzwords at Kulturbrauerei is a unique experience: I’ve worshipped DJs before in most of the rooms where I now had the chance to listen to a broad series of interesting talks with just the right mixture of tech details and concept overviews! There was a very communicative and open-minded atmosphere among the 600 attendees. The sun in the sky and the beer benches in the yard invited the participants to have professional as well as social discussions.
It all started off Sunday afternoon with 150 people attending the Barcamp, participating in 16 different 30-minute talks. Topics reached from visualization art (www.inconvergent.net) over general discussions about probabilistic sublinear data structures to the t-digest algorithm to estimate quantiles (https://github.com/tdunning/t-digest).
Both keynotes dealt with important concepts of today’s software development. The keynote of Ralf Herbrich, Director of Machine Learning at Amazon, on the first day of the conference depicted the time it takes to transfer technology in practice: It took him and his team at Microsoft nearly two years for transferring a new algorithm to improve the ranking of players and match making in online games. Besides the necessity to not only sell but evangelize an idea to make it known and used, I learned that playing video games is an important part of technological evolution.
The second keynote, held by Katherine Daniels, operations engineer at GameChanger.io, emphasized the need for team interaction within software development, especially between the dev and the ops team – which most companies try to solve by adding a “devops” team. I really liked her sarcastic comments about how useful it is to add a devops team to make the dev and the ops team talk to each other. Katherine’s attitude of devops being a mindset of empathy between system administrators and developers rather than a position of communication facilitation inspired me! And even more so since I admittedly could easily identify myself as one of the developers she described:
Besides this second keynote my favorite talks were the hands-on session on CQL and data modeling with Cassandra by Jonathan Ellis, project chair of Apache Cassandra. The corresponding blog post can be found here: http://planetcassandra.org/blog/post/python-driver-overview-using-twissandra/. I also really liked the suggestion on randomizing input data to unit tests by Dawid Weiss, Co-owner of Carrot Search s.c. in order to cope with the real “border” of your code:
I was eager to hear Ted Dunning, Chief Application Architect at MapR and thus partner of my company comSysto, talk about how to use neural networks to detect anomalies to catch live problems early and help any CEO sleep better through the night.
The conference ended for me with a day-long hackathon on Data Stream Mining at Retresco GmbH. We were 20 motivated hackers implementing the heavy hitters and count min sketch algorithm in the programming language of our choice in order to then process and possibly visualize a stream of 8.6 million tweets. The python solution me and my peer (@whitenoise) produced can be found on github (https://github.com/lfoppiano/bbhack-2014). With the focus on getting things done we managed to implement the two algorithms (without much regard for performance) and get a visualization based on matplotlib (even if it isn’t pretty) in six hours.
In summary, the Berlin Buzzwords was a valuable experience worth repeating.