Where I ramble about DevOps -> Agile -> Data-driven decisions

Jez Humble has a lot of really interesting ideas on how companies deliver value to their customers.  Let's start with that.  I've now probably spent a good 15-20 hours just listening/watching his talks (not to start a JH fanclub or anything...he just ties together the work of many into something that makes sense to me where I am today).

Over the last year of this 'DevOps' journey, here is how my research/learning has progressed:
  1. Do monitoring stuff, start to look at 'application health'
  2. Realize there are more important things that need attention, so I give them attention (bottlenecks)
  3. DevOps Demystified w. Ben Rockwood, mind expands.  Understanding 'why' is super important.  Understanding 'flow' is super important.
  4. Analyze the flow! (build/deploy automation & test automation came up as candidates)  
  5. Fixed build/deploy by automating it!  In progress!
  6. Test automation! In progress!
  7. Read 'Continuous delivery', mind expands, yay!  We are on the right track!
  8. Got the team to attend DevOpsDays 2015 Toronto, realized we are not alone, realized we are on the right track.
  9. Listened to what the community is saying about data-driven decisions, culture change, etc
  10. Realize that all of these ideals cannot be driven from the bottom up
  11. Review and rejig: get middle management re-focused & back on the same page (this was...dicey...from the bottom up)
  12. Realize that all of these ideals MUST be driven from the TOP down
  13. Uhhh....I'm an admin. (yes, there are things I can do, and am doing, e.g. lead by example, fix one small thing per day, encourage questions, ask questions, etc)
  14. Heard a talk Jez Humble gave on Lean Enterprise - referenced Ronny Kohavi's (et al.) paper
  15. Had a conversation w. one of the managers about how this whole devops thing is futile without a change in business decisions
  16. Read Mr. Kohavi's (et al.) paper: http://ai.stanford.edu/~ronnyk/ExPThinkWeek2009Public.pdf
  17. Head exploded.
Our DevOps journey started out on the very technical side of things (traditional monitoring) and turned into a discussion on how we decide that something has business value, and the ramifications that has on the rest of your flow.  So when someone says 'analyze your flow', you don't just assume that "the business" is good to go and leave them out of the flow analysis.  To be fair, it wasn't obvious until I'd spent a year absorbing all this stuff, but it's pretty clear in hindsight.

Then I was introduced to agile (coming from Ops, was not familiar) - agile is meant to bring the devs and business closer together.  So DevOps brings devs and ops closer, and agile brings business and devs closer.  What wraps the whole team together?

Maybe an example will clarify.  You want to re-do your product because it's not selling.  So you contact a few customers, get some ideas, build a wall of stickies for stories and whatnot, then say 'ok IT team, build this'.  Great!  But a few questions first:
  • This is essentially waterfall, are we not agile?
  • Are you certain this will work?
  • Are you sure you need everything fully functional?  That's a lot of work.
  • Have you proven that all these ideas have true value?  How have you proven this?
  • Are you aware that decisions without data are statistically proven to be wrong more than 50% of the time? (it's more like 70-90%)
  • Have you read the work of Mr. Kohavi et al.?
(I am not bashing the effort behind someone's work, just applying constructive criticism)

This is super rambly, but what I'm getting at is: 
You can be super DevOpsey and Agiley but still get nowhere.  Does it matter that we can do Continuous Delivery if we are uncertain that we are continuously delivering value?  Do we even need to take the effort of completely re-doing our entire dev/ops processes?  What if we just scaled it down to small prototype projects?  What if we could quickly cancel a project because we know it's a flop before we invest a ton of time?  How do those questions change your approach to "DevOps"?

If anyone is on their DevOps journey and has not considered how your company actually decides to do something, read the paper.  If you're just interested in the idea of 'controlled experimentation', read the paper.

We'll see how this pans out, but I definitely feel that this is the missing link we've been looking for.

Comments

Popular posts from this blog

DFSR - eventid 4312 - replication just won't work

Fixing duplicate SPNs (service principal name)

Logstash to Nagios - alerting based on Windows Event ID