More on the talk after the jump:
In my talk I introduced Snowplow briefly, then showed how our loosely-coupled architecture had allowed us to introduce Scala gradually, starting with our Scalding-based enrichment process for Hadoop, and ramping up a few gears now with our all-Scala Kinesis-based process.
I dived into some detail on the “secret sauce” of our Scala data pipeline, particularly the Scalaz Validation, which is our key data processing primitive, plus our great experiences with Specs2 data tables and ScalaCheck; I also talked briefly about how we handle build and deployment at Snowplow. I also gave a brief introduction to some of the standalone Scala projects we have open-sourced to date in our Snowplow organization in GitHub (such as the referer-parser project).
Finally I rounded off by discussing our roadmap, focusing particularly on our vision around the unified log concept and Kinesis, and gave a sneak preview of our plans to re-implement our Tracker Protocol around JSON Schema.
Please check out the slides for more:
Also Skills Matter have now made the video available, check it out!
One of my favourite things about Snowplow technology talks is the chance to get feedback on our architecture and roadmap, as well as hear about others’ related experiences.
It was great to hear about Graham Tackley’s experiences with JSON Schema and Elasticsearch at the Guardian, and Volker Pacher’s work implementing Neo4J at Shutl. We also got a great suggestion from Ian Morgan at Can Factory, around the need to parallelize our individual enrichments as we start to add more blocking, time-consuming enrichments like currency conversions; this will be particularly important to minimize lag in our new Kinesis-based enrichment flow.
So a great meetup, many thanks again to Andy Hicks, LSUG and Skills Matter for arranging. And as a final note - if you are around in London and would like to grab a coffee to talk Scala, Snowplow and data - do get in touch!