DevOps - Rethinking IT Operations
Large cloud providers approach IT operations very different from the traditional IT ops model. DevOps is about applying agile tools & processes on infrastructure process but it’s more than that, its a culture.
DevOps is complimented by the Pace Layering concept - ensuring the right processes and focus on each layer of the Pace model. Current approach are based on silo-based efficiency, not necessarily end-to-end effectiveness. Moving to DevOps and Pace Layer may well be more expensive than the silo-based approach so you need to clearly understand the business drivers for the change.
DevOps reconciles the more frequent release/development cycles with the structures of operations (e.g. change management lead times). Addressing the speed mismatch.
Great example of DevOps - Facebook grades all of its developers; people with more stars have their changes go in quicker than others with few stars. This builds a culture of trust between developers and operations staff. Facebook works on one major release on one day each week, does enhancements for 3 days and doesn’t put in any change on Friday/weekend unless they are critical.
Co-location of teams was certainly one of the major themes identified by workshop participants. Clarity of roles, expectations and no surprises are also key.
Outsourcing can make DevOps more difficult however it does come down to the relationship and contracts with outsource providers.
Whilst not completely incompatible there are often issues with the DevOps practices in a pure ITIL environment - that scenario requires quite a bit of careful integration to make it successful. As with all changes there needs to be a strong focus on building the right culture.
I think overall we’ve probably taken some steps in the DevOps direction and one of the things I’d expect to come out the application strategy work we have underway would be some statements around where/how far down that road we progress.