Thursday, February 14, 2013

Software Development Seminar: Continuous Delivery - to have or have not



A link to all presentations is below.

Damon Poole (@damonpoole), Chief Agilist at Eliasen Group set the stage for MassTLC’s software development meeting last night at VMWare in Cambridge with an overview of Continuous Delivery (CD) and the tools and strategies needed to move to CD to make more money. He reminded us that we should all be measured on business value delivery not on how many lines of code are written or how many bugs are fixed.  Our ability to deliver value by bringing ideas to market faster and reducing cycle time is crucial.  

Next, Aaron White (@aaronwhite) from Boundless, a start-up in Boston that makes textbooks free online for college students, discussed their implementation of the entire suite of agile strategies, from CD to Continuous Integration to Kanban.  They average 10+ deployments per day and every team member takes ownership from conception to deployment including development, test, QA and their also considering customer support.  To support this they widely use automated testing. They built an internal dashboard to manage their change log, which is crucial to managing what’s going out the door. He also highly recommends Trello as a helpful Kanban project management tool. 



Chad Pytel (@cpytel) from ThoughtBot, a consulting shop in Boston that builds web and mobile apps on Ruby, has two roles for employees at his company: either designer or developer. There is no QA stage as developers have to write their own tests which is critical to the success of CD. They deliver daily but do weekly retrospectives.  They built their own project management tool that is hugely successful for planning time, setting schedules and priorities. Of note, they have a great 3 month training and mentoring program in place to find top talent!

Elias Torres (@eliast), joined us from Hubspot.  At $65M in revenue, they only have 45 engineers and work in small 3 person teams. Each team defines their own process for scheduling & prioritizing engineers and own their products from code to operations. They seem to win the prize for most deployments as they deploy 100+ times per day. Elias did note that going fast is great but it also has its price.  Quote “We break shit all the time - but we also fix things very quickly.” The amount of change between deployments is so small they can either roll back to the last deploy or fix very quickly.  They have also found that sometimes the amount of change can be difficult for customers.  Communications with the rest of the company and their customers is critical. What keeps Elias up at night? If they slow down, the fixes will all be bigger and more could be broken. Some of the challenges are using 1500 Amazon servers which can cause complications. “We focus on developer productivity. We automate everything except writing new code,” concludes Elias. 

Following the presentations, it was a very active discussion with the audience.  Thanks to all who participated. As it was a hot topic of the night, look for a follow-on session on the challenge of Automated Testing.  Interested in being involved?  Shoot me a note at Christine@masstlc.org.  Also, don’t forget to look for our roundtable discussion in April on Managing Distributed Agile Teams.


Presentations:

Thanks to our sponsors:  Constant Contact, DYN, Rally Software, Verivo and VMWare and to VMWare for hosting!


Christine Nolan
@c3nolan
Christine@masstlc.org