From time-to-time, I do a podcast on the topic of Agile. Here’s one in that series that’s worth calling out as it crosses several lines of interest, namely Agile development, IT Management, and even cloud computing. It centers around a discussion with Andrew Shafer from over at Reductive Labs and, of course, Israel Gat. Here’s the cross-post from The Agile Executive Blog:
To listen to this podcast, download the podcast directly, subscribe to the blog/podcast feed in iTunes (or whatever), or click play below to hear it:
As Israel alluded to last week, in this episode of the Agile Executive podcast, Israel and I talk with Reductive Lab’s Andrew Shafer. Put broadly, the topic is “Agile Infrastructure,” which kind of boils down to the connection between Agile development and the IT department, esp. in trying to get IT to be Agile itself. Here are some, admittedly, poor notes from the show:
After some brief introductory stuff, Andrew launches off: traditional operations has built up a resistance to change. “Change is what enables the business.” There’s an interesting discussion here of operations and infrastructure concerns being “non-functional requirements,” which are sort of second class citizens in some Agile practice.
Starting around this point, Andrew starts referring to talks from the recent Velocity conference. There’s special mention of the John Aspaw talk.
I ask Andrew and Israel where, beyond web companies, they see these practices happening or finding interest. Andrew admits that its only web companies that he sees applying this thinking, but analogizes it to early Agile, XP in particular, which had a small, narrow focus at first and then spread over 10 odd years to where it (Agile) is today. Israel weighs in with an example from a few years ago in the financial sector
Having talked about these ideas in abstract, we talk about some of the practices themselves:
- as mentioned about, treating your infrastructure like source code – something you can rebuild on-demand.
- automate your infrastructure – from bare metal to running services.
- capacity planning, but better, management and acquisition – e.g., rebuilding 60 machines from metal to production in a few hours at Digg.com vs. two full days or work.
Israel asks, is ITIL for the data-center like water-fall for development? Both Andrew and I weight in on how much water-fall you can buy into, making analogues to eat-the-whole-pie RUP. This also recalls a conversation I had on another podcast, the IT Management & Cloud podcast with Rob England, aka, The IT Skeptic, on the topic of CMDBs and ITIL.
Disclosure: Reductive Labs is a client.
We are a SaaS software developer. What agile dev teams are discovering is that they can assume more risk and increase acceleration of feature development by combining agile development practices with agile application management. We push a new feature release every week. When we do, we monitor it during and immediately after the push looking for degradation in performance, an increase in errors, or other serious issues which could affect user experience. In many cases we can determine in a matter of minutes if the new release can stay in production or needs to be rolled back to last version. We are able to reduce our testing and increase our pace of new feature development at the same time. Our head of engineering has posted a blog on our approach at this link. http://railslab.newrelic.com/2009/03/06/how-perfo…