#GlueCon 2014: Keynote: The Parallel Universes of DevOps and Cloud Developers – Donnie Berkholz, RedMonk
Keynote: The Parallel Universes of DevOps and Cloud Developers – Donnie Berkholz, RedMonk
- Background as a drug discovery scientist
- Most attendees of DevOps conferences are ops, not developers
- Other conferences have lots of developers, fewer ops
- The result is a community failing to communicate and are often silo’ed out
- DevOps is:
- 1. Cultural – people, not tools
- Worked on Gentoo for some time and found that continuous delivery was happening, allowing code to get to developers within an hour
- Open – open source, open APIs – no standards bodies (that enable enterprise software companies to compete by slowing things down)
- 2. Infrastructure as Code – the idea is old (CF Engine shipped in 1993), but it gained traction as a result of the cloud
- Puppet and Chef emerged as CFE was hard – introduced Ruby to ops (and now Go is emerging)
- Open
- 3. Monitoring
- Nagios the accepted standard as it is open – released in 1999
- Other options: graphite (great store, not great interface)
- Now we have a movement of log analysis to help bring context and machine learning to monitoring due to the large infrastructure
- 1. Cultural – people, not tools
- The best Ops Experience wins (as a funnel):
- 1. Can’t find download
- 2. I need to register?
- 3. Installation too hard
- 4. Poor development experience
- 5. No community of users like my
- 6. These are your users
- CoreOS and other projects work to optimize the experience of a specific group
- Moving from runtime experience (config) to the developer experience (integration, glue)
- Datastores focus on the developer experience (“who cares if we lose some data, we’ll restore from backup”) or the production experience (“it should never lose data”)
- Developers are now looking at outsourcing to DB-as-a-Service vendors
- Ops look at DBs that won’t lose data, reliable, easy-to-install and manage
- The DBs that have been around for a while, they often flip (devs now want pg vs. ops that want MySQL)
- Mongo is in over half of enterprise/JavaEE shops due to the developer experience
- There is DevOps, Cloud, and an intersection of both (“Bay Area startups” and like-minded)
- The intersection is a result of lack of inertia preventing the use of both that other companies/enterprises experience
- It is only changing if companies choose to shift money from IT staff to cloud resources
- China Mieville – “The Ciy & The City” – illustrates the choices made within each city without collaborating/sharing across cities
- Developer priorities for 2014-2015 is: Java 8 (35%), Coninuous Delivery/Deploy (18%), Non-Java programming (15%), …, DevOps (7%)
- There is a disconnect between delivery/deploy and DevOps concepts with developers
- Don’t call everything DevOps, as you silo out the DevOps team from everything else
- Worldview: Code-centric vs. System-centric
- It is about barriers to entry: how easy do you make it for people to work together
- Open, Expressiveness (reduces bugs), and DSLs
- We need to unite at Application-centric: Ops move up a level, Devs move down a level