Madrid ClickHouse Meetup Report


April 10, 2019

The Madrid ClickHouse Meetup is over. Having attended in person I can report it was excellent: well-organized with great content. The meetup venue was the Google for Startups Campus, which has a comfortable auditorium capable of holding over 100 attendees. Between the venue and quality of the talks, it was more like a mini-conference than a meetup.

Javi Santana organized things and did a wonderful job. Among other things he coupled the ClickHouse meetup with a data science meetup earlier in the afternoon. Attendance was great for both events—every seat was taken for the ClickHouse talks.

Speaking of presentations, these were universally good. Here’s a short list of topics and a few remarks on each.

  • ClickHouse introduction (Alexander Zaitsev, Altinity) — Alexander is our CTO and did an overview talk on ClickHouse that included the new Kubernetes operator support that we just announced. This was a nice start as many attendees had data science background but were new to ClickHouse itself.

  • ClickHouse 2019 new features (Alexey Milovidov, Yandex) — Alexey is the main committer for ClickHouse and presented an overview of his favorite new features in ClickHouse. These included several new index types, LowCardinality datatype, and HDFS import/export to name but a few. The slides have many more features but these were among my favorites.

  • From legacy to ClickHouse (Iago Enriquez, Idealista) — Iago describe how Idealista runs analytics on real estate markets. There was an especially interesting description of using Catboost models for data preparation–the models can impute missing values to enable more accurate analyses. It nicely demonstrated the value of combining models and data in ClickHouse. This is a topic you’ll hear about in future on this blog, as we are also quite interested in the subject at Altinity.

  • 1027 predictive models in 10 seconds (David Pardo Villaverde, Corunet) — David showed how to use ClickHouse to generate parameters for running machine learning models.  It’s a nice example of using ClickHouse for data exploration to choose and configure models. This is another state-of-the-art technique for managing analysis of large data sets.

  • Shipping Data from Postgres to Clickhouse (Murat Kabilov, Adjust) – Murat described a new project he has developed to replicate data automatically from PostgreSQL to ClickHouse using logical replication.  There were numerous questions about this as PostgreSQL is quite popular. The project is new but quite promising.

Overall the talks showed imaginative uses of ClickHouse. Attendees came from the Madrid area of course as well as further afield in Spain. I also met people who flew in from the UK, Switzerland, and Germany, not to mention Russia where Alexander Zaitsev and Alexey Milovidov live.

I would like to thank both Javi Santana as well as all of the speakers for a great event. Alexander Zaitsev and Valery Tkachenko from Altinity put in a lot of behind-the-scenes work to schedule talks and ensure there was abundant food and drink. Finally, I would like to thank the audience for their attention and enthusiastic participation.

The Madrid ClickHouse Meetup was great. We’re already looking forward to future events in Spain. See you soon!



Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.