A Developer-first Approach to Kubernetes + Autopsy of a Kubernaughty: Disk IO
Details
Join us in March at the Google campus for 2 presentations – "A Developer-first Approach to Kubernetes" with Manish Vachharajani from Unbounded Systems + "Autopsy of a Kubernaughty: Disk IO" with Jesse Noller from Microsoft.
---
AGENDA
[06:00 - 06:30 PM] Dinner/Drinks (provided by Microsoft Azure) + Beer (CNCF)
[06:30 - 07:10 PM] A Developer-first Approach to Kubernetes (Manish Vachharajani, Unbounded Systems)
Developer-centric tools and technologies, like Netlify, Heroku, or Zeit offer an incredible developer experience (DX), but they force developers into a simplistic one-size fits all approach with no visibility into infrastructure for debugging and performance tuning. On the other hand Kubernetes offers a rich operational experience, with fully customizable deployment, sophisticated monitoring, and visibility into all layers of the system if need be.
Manish will cover how Adapt.js, gives developers a world class DX while allowing them to customize their infrastructure for local development, testing, CI, and even production using technologies they already understand. When combined with Kubernetes as a deployment target, it lets developers build what they need, while allowing platform teams to build in the security, availability monitoring, and other features needed for a world class production environment.
BIO: Manish is currently Founder at Unbounded Systems, a startup working to provide a best-in-class developer experience for cloud-based software, yet still preserve the flexibility and customizability of rolling your own infrastructure. Prior to Unbounded, Manish was a founder at LineRate Systems, a high-performance software-networking startup (sold to F5 in 2013), which in turn was based on research he and his lab did at the University of Colorado.
[07:10 - 07:50 PM] Autopsy of a Kubernaughty: Disk IO (Jesse Noller, Microsoft Azure)
Kubernetes adoption is increasing faster than any of us could have imagined. Unfortunately with this Cambrian explosion of users, tools, new approaches and more we may have forgotten that these large, complex systems are exactly that. With the container native stack existing in userspace and “software defined” you must be able to correlate and analyze failures at the host level – those failures (say, disk latency spikes) have a butterfly effect as it moves up the stack through the container runtime, SDN/CNI – leading to workload failures and overprovisioning of your IaaS. This talk will examine specific `Kubernaughties` around Disk latency, IO and throttling.
BIO: Jesse (@jessenoller on twitter) is a Principal Program Manager at Microsoft on the Azure Kubernetes Service. In the past he’s worked for other Cloud Weapons Dealers, helped build globally distributed filesystems, lead engineering teams – his background is in distributed systems/systems engineering. He lives in Colorado with his 3 dogs and is trying to build a small farm to leave tech because computers.
[07:50 - 08:00 PM] Stick around, socialize & network!