from dev and ops to devops - reconfiguring the plane in flight
Post on 05-Jul-2015
341 Views
Preview:
DESCRIPTION
TRANSCRIPT
From Dev and Ops to DevOps
“reconfiguring the plane while flying”
Mike Wessling
Lead Nerd @Bitbrains
Who is Bitbrains
• Bitbrains provides managed hosting for medium to large enterprises on our own shared platform. – Mainly for financials
– Often HPC platforms and other odd ducks.
• From Day Zero to the Bitter End™– We get involved as early as possible in projects.
– Prefer iterative design approach.• Start building early
• Think, build, test, review, repeat.
• 35 bitbrainers.
Old Organization
Fairly traditional along life cycle steps:• Project engineering (customer+platform)
– Presales consultancy– Proof-of-Concepts– Design– Implementation– Big Changes
• Operations (customer+platform)– Maintaining status Quo– Changes– Maintenance– User requests
• Separated by the big Ops Hand-over– 47 checks long. Some checkboxes are big..
Why Change?
• External influences:– Customers using agile project approaches.
• Even the big ones. • Traditional Waterfall doesn’t deliver.
– Customers want “try before you buy”,– Customers want “pay as you go”.– Applications getting bigger and more complex. – Growth!!
• Internal influences:– The organization didn’t reflect how things worked,– Projects rarely end… – And never get handed-over completely– Customer like the TLC they get in the project phase.
• Responsive and flexible. • SLAs and processes are less important• Willing to pay for it.
– 1 project = 1 engineer is vulnerable • Especially when running at 100%+ capacity
Result: Out of control in 2013
• Still delivering happy customers but– Project engineers overstretched
• Handovers suffered. • No time to make life easier.
– Customers stuck in project phase– Operations struggling with
• Partially handed-over customers, • Complex customers without history and experience• Customers not used to more rigorous processes• Out of date documentation.
– Internal improvements not happing– Platform support stuck between engineering groups.
• No time to grow• Organization can’t scale any further this way. • Engineers working on 20 or more projects
So Now What??
• Need to scale the teams
• Need to reduce pressure
• Need to get the process back in control
• Need to create space to grow.
• Need to get rid of the handovers
• Need to spend time on internal improvements
Lets try this DevOps Thingy
• Seems to solve key problems
– No to hand-overs
– Deals with the ‘Never ending’ projects
– Shared responsibility between Project and Ops.
• We are already doing it
– But Implicit, unstructured & out-of-process
• In our DNA – it is how we started
– So how to scale this….
DevOps? Ops ok But Dev?
In our enterprisey world:
• Read “Dev” as the team ”Specifying, designing and building the service.”
– Not just the “code-monkeys”
– “Coding the stack”
• Read “Ops” as the team “Running and delivering the service”
1st step: Creating Time and Space
• Nothing is going to change otherwise– Customer projects are always priority– Customer questions and requests are second
Approach:• Take work of the plates
– Outsource well defined projects. – Verify/renegotiate target dates.. How hard are they?– Teach Sales not to say yes to all dates. – Cancel/Delay projects if needed.
Enter KanBan
• Start with project engineering – Mature self-aware group – Role model for other engineers
• Start with a good coach (!)– External/neutral – Group will rebel or demotivated at some point
• Requires somebody who stands above the process
• KanBan will expose problems– Be prepared to deal.
• Look for any benefit– Being able to share the work across the engineers for example
• DON’T GET HUNG UP ON TOOLS..– Not important. – Distract – Force a way of working
Result
• Project Engineers happier– In control of work
– Less context switching
– Able to share the workload• New engineer productive after 3 days.
• Project Manager happier– Better overview
– No need to ask all the time
– Aware of the scary pile of work in the queues
• Operations eager to switch to KanBan
Next: Operations
• Bigger challenge– Vast majority of work are tickets
– Any projects are individuals doing their best
– More junior group
• First Results:– Insight:
• many tickets were parked– Returned to queue.
– Bad stats – Service management unhappy
• Lots of little side projects
Operations continued
• Insight:
– Need to add long term and short term engineers
• Short term to clear the heap.
• Long term to stay on top and have room.
• Creation of Platform team
– Different beast with different focus
– Big projects and operational role. (very DevOpsy)
towards: Dev+Ops = DevOps
• Working towards close alignment of processes
– Same Queues
– Same Definitions of Work
• Already work can travel across teams
• Coordinated Set of priorities across teams
– Service management plays key role
– Currently 1 week window, extending to 2 and 4 weeks.
The Hand-over Elephant
• The big divider between Dev and Ops
• Big To-Do list.
– Big block of work after the interesting bits are finished
• Used by Project engineering as buffer time.
• If Ops accept they are stuck with the customer
The Hand-over needs TO GO
But how?
• Engineers took as step back and looked.
• Hand-over must be continuous process.
– Stuff gets added and changed all the time
– Different bits of the environment are in different stages
• Integrated in the definitions of work.
Switch to DevOps Mode
• Flip the engineering teams from a Life Cycle organization to DevOps teams
– Won’t be using DevOps as name
– Teams are end-to-end responsible for a set of environments grouped by common factor and single board.
– Should be a small natural event.
– Number of teams: depends
The End
Questions??
top related