Together with my colleagues from Coolblue I visited the Velocity 2015 conference this October in Amsterdam. Velocity is a yearly event about web operations, end-to-end optimisations, ubiquitous delivery and performance. This year the theme was ‘build resilient systems at scale’. As you can imagine this is pretty relevant topic for a fast growing company like Coolblue :).
It was a great and useful conference, I liked it a lot. O’Reilly organises the conference and they always do a great job in making the slide decks available. You can find them all here. Going through all the presentations will take you quite some time. So here are a few of my favorite topics and talks.
There were several some talks about DevOps related topics. Many from companies that have introduced DevOps methods in their way of working, the failures and successes they have had with it. There were talks for example from Marks & Spencer (M&S), Etsy, Bol.com and the Financial Times.
Etsy talked about how they made sure software went properly to production. Their definition of ‘production’ was: 1. The software is monitored, 2. Someone is responsible for it, 3. It is getting live traffic, 4. It’s being maintained. They explained that combining different disciplines in one team (especially infra and development), at least for certain periods (1-6 months) brought them quite some success. Etsy also emphasised the value of monitoring and dashboarding the performance of the systems you make. M&S and the Financial Times came to that same conclusion. M&S has one simple rule for that in their DODs “Don’t get slower” and they track that continuously.
Sarah Wells, from the Financial Times also discussed monitoring extensively. She warned the audience about the overload of error messages you might get when implementing microservices. Implementing proper monitoring and dashboards saved the Financial Times from an overstressed IT services team.
A hot topic at Velocity was the best way to go to the cloud. Mozilla had a nice story in particular on that topic. Because they had to stop their contract with the DataCenter they were almost forced to make that step very quickly. They discussed their current development process for infra structural changes. This is very similar to their software development process. At the moment they are completely running in different clouds and are very happy with that.
Another cloud story was from the government of the UK, the Cabinet office . Like Mozilla, they also run most of their software in the cloud, also in different clouds. Their experience after two years was that it is very important that certain cloud solutions, really work well for the users (mostly developers in their definition). They advised not to wait to start with clouds while hoping the landscape of providers will become more clear. They saw still so many changes lately with all cloud providers, that their conclusion was that the landscape will remain dynamic for the foreseeable future.
Optimising Teams in a Distributed World
One talk I enjoyed in particular was from Mike Amundsen (API academy). His topic was on how scaling software teams works. He talked about “Dunbars numbers”. These numbers are 5, 15, 35 and 150 and relate to the scale of human networks. When talking about (professional) relationships it might be wise to keep these numbers in mind. In groups of around 5 persons you can for example have an active relationship with everyone. Meaning you can keep track of the well-being of your teammates, you know more or less what keeps them busy and you’ll probably have at least daily one conversation with all of them. In groups of 15 people this is a lot more difficult. This is why scrum teams with more that 7 people are likely to function slower than teams with 3-6 people.
I hope this post gave you some insight about what is to be found at a Velocity conference. If you want to know more about a subject, have a look at the Velocity 2015 website.