EMC World 2015 – DevOps Day – Live Blog

This is a live blog of the whole day broken out by section. I’ll add all of the talks by title at the top – if you see something like, just scan down to that section. I thought of doing slide pictures but this will be posted on YouTube later if you want that.

  • Intro
  • LinkedIn – Shwetabh Mittal – Running & Growing an at-scale social media platform
  • Steve Marx – Dropbox – “It’s complicated: the user/admin/developer/platform relationship”
  • Nicole Forsgen, PhD – Chef Software – “DevOps: Next” – Director, Organizational Performance & Analytics
  • Docker 101 – Jonas Rosland – @jonasrosland
  • The Post-Cloud – Where Google, DevOps, and Docker Converge, Nick Weaver – Intel SDI
  • Competing with Software – It Takes A Platform – Pivotal Exec
Note: these notes are not necessarily comprehensive….what I found interesting as it went along.

Intro

  • Art Min – CTO? of EMC Code
  • Hash tag – #DevOpsEMC
  • Thank you to a ton of folks – EMC Code Team – Cef, Mooo, VMware, Pivotal, Intel, LinkedIn, Dropbox
  • Following DevOps day format – series of talk in one room, other room has Open Spaces/Whiteboards/Demos/Giveaways/Happy Hour (6-8 PM)
  • Open format – please be respectful – letting anyone attend even if not paid attendee of EMC World
  • Github = http://emccode.github.io
  • Twitter = @EMCCode

LinkedIn – Shwetabh Mittal – Running & Growing an at-scale social media platform

  • How many social media profiles do you have?
    • Everyone has one, 90% with 2, 60% with 3, 40% with 5, 20% with 5….give or take
      • Huge competition for social media users….it’s something of a zero sum game.
      • Key tenets of a social media platform
      • Users – Sky is the limit…7 billion or so. Want to have millions or billions of users.
    • Key Questions
      • How to retain existing users while adding new users to the platform?
        • Trust – Users entrust you with their data.
        • “Data is king”.
        • Don’t use data in ways you don’t want users finding out.
        • Transparency – don’t sneak in changes to the platform, provide tools for users to understand platform/policies
        • Value – like any product management need to know who your users are
      • How to gain leverage in a resource constrained environment to deliver value at a fast pace
        • Closed source vs. open source
        • Open source is huge source of value for this.
        • Every web scale giant is heavily entrenched in Open Source.
        • Square has a 1/4 million lines of open source code, Facebook has 10 million.
        • LinkedIn has started a lot – Voldemort, a couple Apache products, etc.
      • What is the long term purpose of the platform and how to rally towards the vision?
        • Create the “Economic Graph” – snazzy animated image on the screen.
        • Overall goal = create opportunity for talent at massive scale.

Steve Marx – Dropbox – “It’s complicated: the user/admin/developer/platform relationship”

  • For anyone who knows AJ Kuftic, this guys sounds like him.
  • Most people call me SMARX – blog.smarx.com – @smarx
  • Dropbox Platform
    • Goal: allow developers to integrate Dropbox into their apps (there are more ambitious goals but this talk focuses on this one)
    • Users authorize apps & apps act on behalf of users.
    • Giving Trulio as an example – utility service to allow developers to build text message capabilities into their apps
  • OAuth Love Triangle – three roles – 1) Server (Dropbox), 2) Client (App), 3) Resource Owner (User)
    • About half the audience understand how OAuth works (raising hands)
    • Graphic on screen to show how OAuth works
    • Means you can get errors from the utility service
  • Example errors
    • “My WordPress backup failed” – so…what’s the app?
    • “I got error 414, and the developer said it’s a Dropbox bug” – interesting b/c means user went to the developer
    • “My printer can’t get files from Dropbox” – cool that printers can access Dropbox but don’t rev API support very often. Some printers are still based on version 0 of the Dropbox API
  • Challenge that at times developers want user info that developers could gather (but don’t) and ask Dropbox for it (email addresses, etc.). Interesting challenge…
  • It’s complicateder – Dropbox for business has team admins, Admins can link apps to the whole team, “My shared link broke”
    • Now the user comes to Dropbox with an issue….and it’s not Dropbox issue or developer issue but the local admin
  • Policy
    • App approval – Dropbox approves apps for production use if have more than 100 users (under that can do whatever you want)
    • Apps must have a privacy policy to access Dropbox data
    • Apps mustn’t claim Dropbox involvement – example = “Dropbox Photosop”

Nicole Forsgen, PhD – Chef Software – “DevOps: Next” – Director, Organizational Performance & Analytics

  • Researchers have recently been able to find a link between IT investments and bottom line – but ONLY if accompanied by investments in culture, people, IT and automation, and process.
    • Would like to think this is common sense but flies in the face of research.
  • We’d like to think that IT investments make the difference…and they do but not enough to get a leg up.
    • This is even the case for really advanced, really expensive ERP systems.
    • Even if gives you a competitive advantage doesn’t give you a sustained competitive advantage….so don’t see benefits last for very long.
  • “Points of Parity” to keep up rather than “Points of Differentiation”
  • We call it the Productivity Paradox – need it to stay competitive but won’t let you get ahead.
  • DevOps is very different….what makes it look different? It’s like the LEAN and Toyota movements in the 90’s – shifts the whole system. This is a culture shift and engineering shift that’s fundamentally different.
  • High performing DevOps teams are MUCH more agile
    • 30x more frequent deployments
    • 8000x faster lead times than peers
  • Much more reliable
    • 2x Change Success Rate
    • 12x Faster Mean Time to recovery – MTTR
    • How do you sell DevOps to the business? These are huge numbers.
  • High performing IT organizations
    • 2x more likely to exceed profitability, market share, and productivity goals
    • 50% higher market cap over 3 years
  • 3 building blocks
    • Culture & Community – matters for any organization.
      • Novelty is valued, messengers aren’t shot, constant learning climate (if we don’t learn, we will die), Dev & IT Ops work well together
      • Job Satisfaction is the #1 indicator of organizational performance.
      • What’s next? Empathy (understanding the users), Burnout (preventing it), Community ()
    • Tools & Automation
      • What we know as best things to do: do version control for all artifacts (not just app code but EVERYTHING), automated testing
      • What’s next – Continuous Integration (CI) & Continuous Deployment (CD), Microservices Architecture
    • Practice & Process
      • What do we know? Cooperation between Dev & Ops, Peer-review change approvals are good for notification but don’t help with lowering error account (actually introduces more errors)
      • What’s next? Proactive & strategic planning, Value delivery w/business players, continue on Kaizen path (yeah…..I need to learn more about Kaizen)
  • DevOps Research in 2015 – looking at Security and compliance at velocity, CI and CD, Exec and org support
  • @nicolefv & nicoleforsgren.com

Docker 101 – Jonas Rosland – @jonasrosland

  • He spoke at Varrow Madness! 🙂
  • Around 50% of the audience has played with Docker
  • Launched a bit over 2 years – launched by DotCloud that later changed its name to Docker
  • Introduced during lightning 5 minute talk at PyCon – didn’t even have time to complete the demo – got pushed off stage partway through his demo
  • Then it just took off with a huge organic community push
  • Beginning of docker – why not scale with just load balancers in front of multiple copies?
    • It’s not just scaling but also about “Separation of Concern”
    • Want to be able to update applications individually.
    • This brings us into Microservices (also mentioned in a previous chat).
  • Microservices = the same application but pulled all the pieces apart.
    • This does make the application more complex – it’s a great idea but adds real complexity.
  • Everything is loosely coupled.
  • When want to scale? Still have presentation layer – then scale out the various pieces. Not scale out the whole app with load balancer in front but scale the individual micro service.
  • So how do we deploy it?
    • One app per physical server? No.
    • One app per virtual machine? No.
    • Containers!
  • What do you need to run a Python app? Not really Ubuntu….but really just the relevant Python libraries.
  • Usually have one function per container – if a Python app, that’s it.
  • Containers are separate from each other – showing a picture of a container ship with each container isolated. Doesn’t matter what you put into each container because of that.
  • Traditional Services — Hardware –> OS Kernel –> Multiple Apps on top (why you have conflicts)
  • Inserted a virtualization layer above the hardware – Hardware –> Hypervisor –> lots of OSes –> lots of Apps
  • Containers look like – Hardware –> OS Kernel –> (Apps but really Containers)
  • Isolation is key without needing a Hypervisor
  • Boot2Docker – tool to run Docker on your laptop – http://boot2docker.io
  • Many minimal Linux OS releases to help you run your containers.
  • What about added functionality
  • Summary – Docker can let you run any type of application anywhere (laptop, AWS, vCloud Air, your own datacenter) — anywhere can install a Linux host as well as Microsoft
  • Read this…

The Post-Cloud – Where Google, DevOps, and Docker Converge, Nick Weaver – Intel SDI

  • EMC to VMWare to Intel – SDI-X group
  • Progression of how we produce things over the years – lots of producers and few consumers.
    • Manual labor, slaves, forced labor, etc. – big jump was the Industrial Revolution. Machines allowed creating more things so we consumed more things.
  • Big Switch – fewer producers and many consumers.
  • Jevons’ Paradox: The proposition that as technology progresses, the increase in efficiency with which a resource is used tends to increase (rather than decrease) the rate of consumption of that resource.
  • What does efficiency look like to us? (as technologists)
    • Walking through server changes…..then database changes.
    • Overall things aren’t getting simpler…they’re getting more complex and more powerful.
  • DevOps has been most a cultural change – letting people run more things.
    • DevOps = efficiency in controlling change.
    • That we as technologists can run more stuff.
    • That shouldn’t just let us do more….but also not burn ourselves out.
  • PaaS – what an Apps needs – like IaaS but closer to the point.
    • Great starting point for container integration (Docker, Rocket), cloud-native apps, etc.
  • Pumping Latice – latice.cf
  • Container Benefits
    • Faster lifecycle vs. virtual machines
    • Contains what is running within the OS
    • Ideal for homogenous app stacks on Linux
    • Almost non-evident overhead
    • Containers start REALLY fast – 1 second vs. 1-2 minutes for a VM
    • This is fine if you’re starting at VM that will run for 6 months – it’s big deal if want to spin up something that runs for 8 seconds and is killed.
  • Running out of steam on taking notes on this session (dealing with some scheduling for the coming week).
  • “The Road to Awe” – where Nick would like to see things go…
    • Want to deploy my web services – x86 servers
    • Want my web service deployed on a VM – VMware
    • Want it automagically configured – Puppet, Check, DevOps
    • Want it packaged and released as a service – PaaS (Cloud Foundry)
    • Want it to be immutable images I can…. (all of this is Docker)
      • turn up in a second for a developer to work on
      • do testing cycles
      • turn up for production – Docker
      • want them to be spun down – Mesos, Kubernetes, CF Diego
    • Want it to be managed intelligently and run anywhere –
    • States….where we’re going (also a bit scary to me – Andrew)
      1. Wider telemetry
      2. Google Omega-style decision making for the masses
      3. Brand new disciplines around efficiency
      4. Massive cooperation of cloud domains
    • Post-cloud – just a label for the computing that our grandchildren will take for granted as they provide their own value.

Competing with Software: It Takes a Platform

  • No one is entitled to their business model – showing lots of companies that have exited the S&P 500 that were too big to fail.
  • Everything is technology driven now – financial services (Square), entertainment (Netflix), 
  • @littleidea – You are either building a software business…or you are losing to someone who is.
  • Traditional IT vendors are declining – Cisco, HP, Dell, etc.
  • Consumer expectations are going up…but it’s startups that are not only meeting those expectations but also building those expectations.
  • Changing development and release disciplines
    • Agile/Developer Productivity/TDD
    • Continuous Delivery
    • DevOps
    • Cloud-native App designs
    • Culture/Talent
  • Amazon deploys code very 11 seconds.

Had to pack up after this…the sessions are recorded. The Puppet one was particularly good.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s