- Product scale: Joyent (acquired by Samsung in 2016) offers API-driven cloud IaaS and a unique Docker PaaS from a worldwide footprint of data centers. Joyent’s infrastructure powers apps serving billions of mobile devices and hosting exabytes of object storage data.
- Product momentum: Container Name Service solved 80% of network ingress use cases faster than we could build a full LBaaS, increasing velocity to address gaps like logging, monitoring, and fabric networks (VPCs). We had to ruthlessly prioritize our efforts so that our small team could execute competitively against far larger clouds. I leveraged RICE and critical user journey, among other frameworks to narrow our scope and stay on track.
- Growth multipliers: ContainerPilot and the Autopilot Pattern solidified our product narrative of sci-fi-like developer productivity and joy and created great community response by proposing a new way of thinking about operational challenges and offering working patterns that developers could use everywhere (not just in Joyent’s cloud). I used customer development strategies, treated every meetup like a focus group, tested big ideas with small prototypes, and iterated to insights and growth.
- User experience: Great cloud user experiences start with gitops and continue with a sophisticated web UX to make visible the logical framework of our apps and infrastructure. Previously: building a great reading experience and maximizing ad inventory.
- Documentation: documentation is marketing, developer relations is customer development, and user stories are documentation.
- Sci-fi DevOps
- Lies we tell our code
- Installing Triton Datacenter (Joyent’s competitor to VSphere) in front of a live audience, based on my blog post
- Also: research notes for a talk I was prepping on operational endurance, with a summary of lessons we can learn from aviation endurance flights
- Long ago: Cloud Mafia meetup (2012), Boston Library Consortium Annual Meeting (2007)
Sample technical background
The links below go to items I wrote or products I built, as well as the work of the team I built and mentored as we scaled:
- Range of concepts: from the OS/hardware interface up through distributed application challenges including CRDTs, Lamport clocks, scheduling and resource optimization, orchestration, discovery, persistent storage, secrets management, service mesh1, gitops, infrastructure as code, multi-cloud, hybrid cloud
- Scheduling tools and approaches: Kubernetes2, Nomad, Mesos, serverless
- Core cloud services: virtual machines (competitive research), containers, Docker PaaS3, auto scaling, network ingress, load balancing, API gateways, DNS, logging, monitoring, VPC (competitive research), object storage (competitive research)
- Scale: infrastructure and services hosting applications that served billions of mobile devices, storing exabytes of object storage, for thousands of startup and enterprise IaaS customers.
In addition to the multi-cloud Kubernetes service blueprint linked there, my experience includes deep exploration of internals while investigating the possibility of offering API-compatible service on top of our serverless Docker PaaS. Separately, I must acknowledge the risk of kubesprawl, including the management overhead and excess carbon emissions it can lead to. ↩︎
It is a point of pride for nearly everybody at Joyent that AWS copied the Triton Docker value proposition with their Fargate service. Us ↩︎