This document summarizes the story of Project Bitesize, an internal startup at a company to address issues with their IT operations. It was founded in August 2015 to fix problems like a slow and bureaucratic internal IT, lack of standards, and a 3-year planning horizon. They chose to use containers and open source tools. Over time, Project Bitesize grew to provide a next-generation application platform and global technical community for developers. It aims to transform skills, engage employees, and become a sentient managed service to automate operational responses. The document outlines some of the challenges they faced early on and areas where they are seeking help or involvement from others in the company.
5. A Brief History…
• Internal IT bureaucratic and slow to deliver
• Lack of standards in application design and build
• Spotty adoption of QA and Security standards
• Release process is manual, slow and high-risk
• Absence of collaboration between engineering and ops
• Roadmap planning horizon is 3-years out
A perfect storm is brewing…
5
6. Our Inception Point
• On the precipice of perpetually failing
• Previous initiatives were huge programs
• Developers running their own tooling
• Business headwinds increasing
• New leadership
• End to end review of tech strategy
6
• Appetite to do something different
• Appetite to do something different
• Appetite to do something different
• Appetite to do something different
• Appetite to do something different
• Appetite to do something different
Introducing Project Bitesize…
17. Accidentally
#serverless
17
Developers
“I need a…”
“I want to…”
Platform API
Developer Tooling
Interfaces/Plugins…
Platform Tooling &
Automation
Global Enabling Functions (TechOps, Security, QA…)
Human feedback loops, reinforce system trust model
18. Platform API Abstraction
18
Security Engineer
Network Engineer
Testing Engineer
Project Manager
Architect
Application Engineer
Cloud
Platform
Engineering
Global PMO
Enterprise Architecture
Customer Teams
CISO Group
Global Network
Global QA Group
A Vehicle for Enabling Functions
Application Developer Community
19. DevOps Target For Pearson
19
Global Technical Community
Community
Engagement
Developer
Experience
Next-Generation Application Platform
Best Practice
and Standards
Initiatives
New Technology
Adoption &
Integration
Hackathons,
Brown Bags
and Meetups
“Do we already
have a … I can
contribute to”
20. If This… Then That - The New Operations
20
Things that Happen Desired Response Mechanism to Execute
Stuff People Care About Stuff People Concentrate On
Trigger Systems Workflows
Tooling
“A synthetic transaction is failing”
“I want to deploy my application”
“Our A/B soak completed
successfully”
…
Re-spawn pods, notify owners
Execute deployment process
Converge all production instances to
B deployment
…
21. Cognitive Learning & Decision Making
Systems (e.g. Watson)
Long Term Vision
21
Trigger Systems
Log Analytics, Monitoring, SOC…
Operational Run Books
Codified and On-Demand
Service/App Metadata
Service Registry/Discovery
A Sentient Managed Service..?
23. Areas for Breakout Discussion
23
• How does our Startup manage “hyper-growth”?
• When does DevOps it stop being DevOps?
• How do we recognise productivity gains vs cost savings?
• Skill transformation, how do we engage Pearson’s veterans?
• Setting the maturity bar, how do we qualify Platform suitability?
What is an “end to end stack”?
When you include people, the technology takes a back seat - I had a lot of “yup” moments listening to Jez this morning
So this talk is as much about getting an initiative off the ground as it is about a showcase of cool tech
Ex-Racker
Frustrated Coder
Education Disrupter
Reformed Leader
This…
@chriswiggy
Tell the story of realising that my aspirations to be an engineer lead me to being a leader… I succeed through others
Expectation set - this means I may not have all the technical details some of you may crave!
My mentality - why not?
Pearson PLC was incorporated the same year Samuel Morse invented the code that inherited his name - thats over 170 years ago
Its a job with high moral fibre
Digital transformation to a global education platform
Good place for someone who’s mentality is “why not”!!
How we’re structure
Self service, highly done automation, very approachable service teams; consult and help and optimize and support
Consult vs. open heart surgery
PaaS: continuous delivery pipeline; how do we give developers an 80% experience; runtimes and containers
VMs -> containers is a couple lines of config changes
1 person is COO/CIO/CTO
Team of SVPs who runs tech ops (enterprise IT)
VP hosting/delivery; I work for him; all next generation infrastructure comes from my team
8 engineers across US and UK; we spend a lot of
Insert large enterprise logo here right?
Umbrella company: Used to buy companies: production company for Baywatch!
Pay off 20-30 years of M&A debt
7 different Ops models: traditional Ops problems, but supersized into AWS (running on unreliable commodity gear: not “HA, won’t go down”): driven by a CTO, “move everything to AWS”;
circa 2012: led to backlash to governance and controls; then tried to build competitor (“private cloud”): “enterprise virtualization with API”; fundamentally wasn’t fit for purpose
Walked into Pearson where everyone actively hates IT; “not just passive aggressive disdain, active hate”
Can’t just embed Ops into Dev teams; they’ll just go native
Outline the chat we had that started this all.
Here is how we snatched victory from the jaws of defeat…
From day one of our pilot, we’ve tried to be different… not for the sake of trying to aggravate and upset people, but because we need to break the short circuits happening inside talented people
We’re a startup inside an enterprise… Our job is to disrupt sustainably from the inside.
Why did we get approved?
containerisation: densification of hosting estate
developer productivity
recognition that we needed to innovate more to go digital
This is why we call it Bitesize
In large enterprise, every word gets one shot; if it doesn’t work, and it didn’t work; we already broke private cloud and DevOps
We had turned DevOps in to a justification for another Silo
We found this wasn’t about a resistance to DevOps, it was a resistance to current operations. For us, starting our transformation was about admitting mistakes and re-establishing credibility.
Make the effort and earn trust/respect… you’ll have credit in the bank when you need to spend it.
Example from this week - perf testing
Self-preservation based on fear of failure
If you have a product, who is your customer?
How do you generate developer engagement - who are we competing with?
Developers want
DIY but not do everything
Self-service but do it for me
Don’t get me out of bed at 3am
And I don’t want to wait for any of this
AND I reserve the right to change this at any time…
Water-scrum-fall
Capital asset programmes
PMO
Funding model: traditional Pearson wants a 5 year plan, write of costs against planned benefits; but I don’t know what I’ll be doing in 5 months, let alone 5 years
We’re one of a few strategic programs, with capex and opex, which other leaders must support; we capitalize as much as we can; long term, we’ll need funding from dev groups, right?
Currently, no mechanism to do chargeback/chargethrough
When you’re coming to the party late, you’d better bring something interesting
We need infrastructure as code, easy to change, easy to test, easy to ship
One of the few benefits of being so late to the market is we can skip stuff others tried already
- Config management, state management, metadata fragmentation…
Macro-scale, we’re trying to simplify and standardise Pearson: less natural mutation in containers
We need to globally deploy, a container is a unit of deployment that we can consider a standard as it is reproducible as long as our platform layer is consistent
The affinity with education and OSS around sharing knowledge is undeniable
We’re embracing open source because:
Its cheap
We attract better engineers
It sets expectations about what we need to be capable of internally
The more we share, the more open we become, the more we share problems
Simple goal - standardised delivery pipeline capable of delivering a consistent application binary to any cloud platform anywhere in the world
Automatically…
explain cycle time - didn't want to scare them!!!
DevOps per page 1 of the textbook won’t scale
400+ dev teams != 400 DevOps!!
Our product aims to deliver a no-ops experience to a user, when in reality it couldn’t be further from the truth
What do our people do with the time? Build relationships, reinforce feedback loops with deeper human interaction
DevOps per page 1 of the textbook won’t scale
400+ dev teams != 400 DevOps!!
Our product aims to deliver a no-ops experience to a user, when in reality it couldn’t be further from the truth
What do our people do with the time? Build relationships, reinforce feedback loops with deeper human interaction
simple yaml workflows - > how do we want Pearson/app/platform to respond when X happens… Stackstorm
By looking at trigger and action, we don’t care what tool we need and we can change it with minimal overhead
Designing the abstraction of this is key, simple API/Provider model
The future of operations is more engineering…
Pearson’s motto is “Always Learning” - we’re applying that to our platform, how do we build a learning platform to avoid incidents and points where customers will get a less than desirable experience?