Quick Thoughts on OSCON in London 2016

Docker talk


  • Open Source software isn’t the best business model
  • Containers enable so much for you
  • Open Source software still have many vunerabilities
  • After Open Source, people want Open Gadgets


Moving open data from tech to culture

Rating: 👍🏾👍🏾👍🏾


  • Accountability for business
  • Reduce cost for businesses and government
  • “Not every road will be open”
  • “Releasing your data doesn’t equate to gaining a competitive advantage”
  • Having a business model predicated on selling data becomes like DRM. The cost of selling the data eventually doesn’t become profitable.

Dealing with cultural diversity and internal advocacy within a distributed team

Rating: 👍🏾👍🏾


  • Embrace differences
  • Highlight achievements consistently (weekly, daily, etc)
  • Ask if an org is “rewarding people with opportunity?”
  • Show and tell
  • We have 2 min presentations but how often do we present?

Docker from development to production

Rating: 👍🏾👍🏾👍🏾👍🏾


  • The intersection of on-premise and the cloud: Docker is at the center
  • Docker has the isolation level to satisfy everyone (on-premise and cloud users)
  • The importance of the OS has shrunk
  • WinServer 2k16 has a “nano server” edition?
  • Orchestrators after dust has settled: Mesos, Docker Swarm, Kubernetes
  • When thinking about dev-ops
  • Agility and normalization of infrastructure are important
  • The ratio of dev op engineers to number of machines is in the 100s or thousands now.
  • Docker images reduce the download bandwidth and increase the installation speed
  • Linux kernel isolation
  • cgroups → Control groups (https://en.wikipedia.org/wiki/Cgroups)
  • namespaces -> https://en.wikipedia.org/wiki/Linux_namespaces

Patterns and practices for real-world, functional, event-driven microservices

Rating: 👍🏾👍🏾👍🏾👍🏾


  • A MS should only have 1 purpose to exist
  • MS gives us a more even distribution of complexity
  • Chaos Engineering
  • “Controlled” experiments on a distributed system
  • Jet.com (Walmart bought) uses F# extensively
  • Claim that functional language has helped with MS development
  • Less code and more readable they claim
  • Pattern matching features help reduce code in MS
  • Be functional, treat unit of work as functions
  • Avoid state changes
  • Makes parallelization easier
  • What they use (big proponents of Azure stack)
Jet.com’s stack

Inside the open source sausage factory: Lessons learned from decades selling open source

Rating: 👍🏾👍🏾👍🏾


  • OSS directly competes with many markets
  • OSS promised to end twisting people’s arms to pay for version upgrades
  • Last 10 years have been critical for OSS
  • The OSS dilemma
  • It’s hard to be profitable as a support subscription model
  • GPL license usage is shrinking, Apache-like licenses are growing
  • This makes it harder to monetize
  • Redhat
  • Only company that has managed to make a support a business
  • Option 1: Support disguised as “subscription”
  • Option 2: Proprietary software disguised as “open source/core”
  • Option 3: Completely open/Completely closed
  • Option 4: Don’t sell open source
Why OSS support doesn’t have longevity

The wonderful world of webhooks

Rating: 👍🏾👍🏾


  • Basically a Pub/Sub pattern
  • Use when you want to notify other systems
  • Webhooks are HTTP
  • Usually between 2 server side systems
  • Typical architecture uses a DB or dedicated queue
  • Webhooks used mostly with 3rd party customers but apply to internal systems as well
  • When receiving events don’t validate or process in order to be partial to speed

Monitoring 101: Finding signal in the noise

Rating: 👍🏾👍🏾👍🏾


  • Instrument all things
  • Including tasks and dev environment processes
  • Can you answer the question “Am I losing money?”
  • This is the question teams have
  • No CEO is saying “Your disk usage is too low but your QPS is great”
  • Ask yourself “What should you demand from your tooling?”
  • VMs have a much longer lifetime than containers that run on the VM do
  • Instrumenting all of this helps
  • Move from a host centric model to a service centric model for monitoring
  • This is like shifting from Ptolemy’s geocentric model to a heliocentric model

Building a powerful data tier from open source datastores

Rating: 👍🏾👍🏾👍🏾👍🏾


  • “Polygot” Datastores
  • Why? → hyper-optimzation of verticals
  • What’s in the OSS datastore landscape?
  • Document
  • Key value (in memory vs durable)
  • Search Engines
  • Coordination Services
  • How do you pick datastore?
  • Does it satisfy a business requirement?
  • Does it have a solid community?
  • Does it plug into your stack?
  • How badly broken does is appear?
  • What does it enable?
  • Yelp guys showed how many features on Yelp pages are enabled using Lucene

Beyond open source: The democratization of software

Rating: 👍🏾👍🏾👍🏾


  • Previously the industry championed OOP until it become ingrained into the culture
  • About interchangeable parts → objects, interfaces
  • Inheritance → because we had “closed” systems
  • Now all software is OO
  • Currently OSS is trying to solve democratization
  • “The App store is the revenge of OO”
  • You simply replace the app with a new version
  • It’s closed source, you can’t replace single part”
  • Issue is “No access to source code, but you can replace the runtime binaries”
  • Ballmer → “Free software is cancer”
  • We need isolation
  • This leads to containers and virtual networks
  • What we really want is runtime isolation
  • The cloud is supposed to do this
  • But then we want runtime versioning
  • DRM is about the runtime
  • Git has versioning but no ACL ​

Links to slides

Additional Photos