Director Of Engineering
CurrentAs usage of our Kubernetes based Compute platform at Zendesk has grown, so has the Compute team. Zendesk is now running close to 100% on Kubernetes, and my org is made up of scrum teams spread across the US and Australia, each with its own distinct concern. - Compute Platform team is the evolution of the original Compute team, now focused on building out a new layer on top of our Kubernetes solution, a control plane of control planes if you will, to automate workload placement and workload migration strategies. Our goal is to increase reliability and reduce blast radius through multi-cluster workload splitting & rebalancing, multi-dimension HA, and strategies for regional, AZ, and cluster level failover capabilities. Soon, we'll deliver Blue/Green cluster migrations with canary analysis to Zendesk. - Compute Infrastructure team focuses on automating the build, delivery, maintenance and upkeep of Kubernetes clusters at Zendesk. We have a complicated fleet of clusters, made up of many different host-groups. When rolling out updates to the fleet, we soak all our changes in a single AZ of just one hostgroup, where we run smoke tests and monitor SLOs. We continue through all Hostgroups and AZs, one cluster at a time. In short, our rollouts are complicated. Infrastructure team is building out a system of operators managing the process, orchestrated by Spinnaker pipelines. - Finally, the Compute Accelerate team is focused on elevating the business in 2 ways. First, by applying the features of our platform to balance the competing concerns of resilience, reliability and cloud cost control. Second, by enhancing and extending the features of our platform to provide more power and autonomy to engineers working the Zendesk suite of products. Currently, their focus includes developing a deeper integration between our Kubernetes platform and the Istio service mesh, and developing a more robust horizontal autoscaling capability.