A conference series that matters – From NoSQL to distributed

Since November 2013 I have been attending six conferences of the conference series formerly known as NoSQL matters, and that now runs by the name distributed matters. And you can say I am a fan! The conferences take place in various interesting places of the world, so far they were hosted in Cologne, Berlin, Paris, Dublin and Barcelona. I have not been everywhere, but my favorite city is Barcelona. Beautifully located at the Casa Convalescència, it is fun to attend all the talks on different topics and discuss with the speakers and attendees. So, after being to distributed matters in Barcelona again on November 21, I want to take the chance to wrap up and share some of my impressions.

Why I like the conference series

What I really like about the conferences is that it is still quite small. There is no rushing from talk to talk, nor are there people who just come to give their talk and leave immediately afterwards. This makes it possible to get in touch with the speakers, even the ones with high reputation in the field of modern database technology and distributed systems. So far, speakers from Facebook, MapR, Google, Codecentric and many more companies shared their knowledge and experiences, and were available for discussions during the day or at the get-together.
I attended talks that go beyond the commercial tech-talks that is sometimes heard at bigger conferences. I have my personal highlights for every year:
In 2013 in Barcelona there was a talk by Doug Turnbull where he took up a historical review on the NoSQL movement which inspired us to have a critical look on the NoSQL movement in Germany in a blog post, and later in an article series for the German Java Magazin (also available at the synyx homepage).
The following year, 2014, Ellen Friedman beautifully told a story on the emergence of big data and the usage of time series databases of which we blogged a short recap that became relatively popular.
And this year Pablo Chacin gave a talk on The tale of two microservices, which focuses on the different interpretations on the weakly defined term microservices (which, in my opinion, is comparibaly weak as the term NoSQL).
And I do not forget to mention the talks of Uwe Friedrichsen, who is kind of a permanent guest at the conferences, and whose talks I always enjoy as they reveal intersting insight into his experiences with modern information technology trends.
The new trend is going from distribution to microservices, and microservice is, like NoSQL, a buzzword with a maybe limited survival chance. But I like that the topic is controversially and honestly discussed in the community at the conferences. Often, a clear and honest message is promoted: If you don’t need microservices, don’t do them. And, as pointed out at several occasions, honesty is a valuable good in modern distributed datastore times.

The conference organization

To my opinion, the conferences are always very well organized. From my side a big thanks to Katja Keil and Jana Volkova from the organization team! Well chosen venues and a good organized entrance make participation easy and free the participants to focus on talks and discussions. And also (or almost most) important: There is always very good coffee. 😉 The traditional after-conference get-togethers provide local cuisine (like famous tapas and beer in Barcelona) and promotes further discussions in a calm atmosphere. Keeping the same locations in cultural interesting cities makes it easy and fun to return on a yearly basis, and also to plan elongated weekends with family.

#dmconf15 in Barcelona

A few impressions on this years conference in Barcelona, that took place on 21.11.2015. For the sake of brevity, I focus on the (in my personal opinion!) most interesting talks of the day.
The conference key note was held by K. Hightower from Google. His talk focused on resource usage optimization in virtualized environments. If you virtualize your machine, there is a good chance that you waste resources, what limits the capacity of machines that can be hosted. That is not only a waste of money, but also puts harsh limits on the number of machines. Algorithms that optimally schedule machines in the environment are needed. Hightower showed kubernetes, an open source project that serves the purpose.
B. Samwell, also from Google, gave an introduction to Googles F1, a distributed database build atop Spanner. F1 is a hybrid database that takles replication inside and across datacenters, and combines the scalability of BigTable with the usability of a SQL database. F1 provides scaling without NoSQL and focuses on OLTP and OLAP as well. As F1 resides atop Spanner, snapshot isolation is guaranteed. Samwell warned that migration to F1 might be hard (it took him and his team two years to migrate from a MySQL cluster), and to be aware of hidden actions when you use ORMs.
P. Chacin told the interesting Tale of two Microservices. As the term microservice is rather weakly defined, he looked at microservice from two different philosophies: Those considering it as an architectural style, and from the point of operations. They differ in certain points but also have their similarities, also to the SOA architectural style. Chacin emphasized that no complete decoupling of services is achievable, as their inevitable coupling dependencies in time (synchronization), location (binding), knowledge (schema), centrality (orchestration), context and state (persistence). He underlined again the law of conversion of complexity in software design (see e.g. M. Feathers article), and the role of size and usage of microservices (cp. the blog series by J. Cramon). Talk and mentioned article are definitely recommendable.
Again, U. Friedrichsen was present as speaker in Barcelona. He warned of the way down microservice road towards microservice hell. He shared some of his experiences and pointed out: You can go down this road, but you do not have not! Those that are not able to build up a well-structured monolith might not be able to master a complicated distributed (micro)service architecture. But microservices might be a choice if the time from idea to market is of high business importance, as well as the autonomy of teams. He also emphasized Evans’ Domain Driven Design approach, and the importance to focus on business functionality, not on data. The DRY principle might not be a good idea when applied across process boundaries, evolution of interfaces is usually unavoidable (as business is due to change!). At the end of the talk he stated about consistency: The real world, he said, is either BASE or inconsistent. Worth considering in discussions about ACID in the near future..

Conclusion

I still recommend the conference series to everyone interested in NoSQL and distributed data storage and systems, and I am in good hope that the quality of talks and speakers can be kept up. As long the the number of participants stays relatively small, discussions will keep to be insightful and fruitful. For me it is clear to go by BlindBird to Barcelona again next year. I hope to see some new faces then, as speakers and in the audience! 😉