QCon London - Day One

Some notes and opinions on the talks I attended on the first day of QCon London 2014. A very good first day; looking forward to tomorrow.

“Life, The Universe, and Everything” by Damian Conway #

I always enjoy seeing Damian talk. Interesting, funny - a great way to start the conference.

“Development, Deployment & Collaboration at Etsy” by Daniel Schauenberg #

An interesting overview of how Etsy manage the development and deployment of their website. The main takeaways for me were:

• Monitor and graph everything that changes
• Create dashboards - lots of them
• Stream all your logs and make them available in a web page

Curious that all of Etsy, including their admin views, is just one big PHP application. This tweet says what the rest of us were thinking:

Etsy at #qconlondon make it clear to me why monolithic apps are a dead end. Use microservices for continuous scalable deployments

It’s obviously working out for them, so not quite a dead end, but it’s surely not the future.

“Real Data Science at NASA” by Chris Mattmann #

Interesting talk on some of the challanges faced at NASA and some of the tools they are using to solve them. Was a bit light on technical detail for me - well, in terms of tools and architecture… There was plenty of detail about the projects, and a lot of acronyms, as summed up nicely in another tweet:

A lot of hairy terms flying around at the NASA talk at #qconlondon. I feel like a #java programmer who wandered into a #haskell talk.

“Revealing the Uncommonly Common with ElasticSearch” by Mark Harwood #

I had already seen a version of this talk at the recent ElasticSearch User Group meetup, but was worth seeing again. A great talk on how you can use ElasticSearch to detect anomalies in your data. Uses the upcoming significant_terms aggregationn which should be available in 1.1.

I can think of so many use cases for this - even if it is just for categorising/tagging log files on the fly. Can’t wait to try it out on some real data.

Design Patterns for Large-Scale Real-Time Learning by Sean Owen #

Sean talked about Data Science and Machine Learning in general and what it means, then gave an introduction to the Lambda architecturee and Oryx.

Oryx can be seen as a reference architecture on how to build a simple lambda-like architecture, but going forward it is likely there will be better systems to help achieve this.

Once again, I was hoping for more technical detail in the talk. This was more of an introduction to a topic I’m started to gain an understanding of.

Instrumenting Your Business For Success with DevOps by Robert Benefield #

I thought this talk was going to be about the metrics you should consider collecting to ensure your application/business succeeds, but instead it was more about how to show off DevOps to people in business (managers, etc). So this wasn’t really my cup of tea. Maybe I read the description wrong.

A Brief History of Data by Damien Dallimore #

The solutions track contains talks put on by the sponsors. Some of them sound like they are just sales pitches, but this one looked more interesting than that so I went along.

And it was quite interesting, looking at how data might be changing in the near future (more of it, more variety) and what changes we need to make (for example, the schema needs to be defined at read time, not write).

Then the second half was the sales pitch for Splunk. They do seem to have a good product - similar in some ways to what we are building at work¹. Good to know we are looking at data in the same way.