2nd day in Antwerp where Torsten, Julian, Roland and Fabian are able to listen to awesome university and tools in action talks. Tomorrow on Wednesday the rest of the ConSol posse (Ana, Georgi, Christian and Christoph) arrives in Antwerp to see the 3-day conference part of Devoxx. And still another highlight to come on Devoxx day 2: Roland speaks about “Spicing up JMX with Jolokia” in his tools in action talk scheduled 18:05-18:35 in room 8.
So here is the wrap up of Devoxx Day 2:
Ludwine gave a very brief introduction to Hadoop, MapReduce and Machine Learning. The focus was on the performance issues that Hadoop might have especially with iterative processing that is a very common pattern in Machine Learning.
Spark solves those issues. It consists of a core layer and the additional components Spark SQL, Spark Streaming, MLlib (machine learning) and GraphX (graph).
Spark core offers very fast in memory data processing - i.e. in memory map reduce - and also a much faster map reduce on disk than Hadoop. Ludwine explained some typical use cases of machine learning and how to implement those with the Spark ML Library.
The talk was a crisp introduction to the ML topic and to Spark that made me curious to see what Spark and Spark ML offers in detail.
Today, I saw the largest and most deeply nested Stream API call ever in José Paumard’s session on “Java 8 Streams & Collectors : patterns, performances, parallelization”. The call solved a Scrabble puzzle in a “single line of code”, where “single line” means there is only a single semicolon. Actually, this “line” was about 30 lines long.
As a university example this was very interesting, and it was fun to solve the puzzle. However, I hope nobody would seriously use this in production. Maybe it’s time to update our Sonar Cube and add a rule for the maximum number of characters between two semicolons.