Cloud Native Computing Foundation KubeCon + CloudNativeCon North America 2016 (Seattle), 16 Nov 2016

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Keynote: Open Innovation: Kubernetes on Google Cloud by Aparna Sinha

Description

Keynote: Open Innovation: Kubernetes on Google Cloud - Aparna Sinha, Senior Product Manager for Kubernetes and Google Container Engine, Google

Aparna will speak about Kubernetes humble beginnings when it was just an idea at Google, to its global growth of contributors, partners, and users. She'll share where the project is heading and highlight how Google is championing for user choice.

About Aparna Sinha
Aparna Sinha leads the product management team at Google for Kubernetes. Prior to Google, Aparna has worked in enterprise software for 15+ years. She was previously Director of Product Management for NetApp’s manageability software where she also led development of storage plugins for Oracle, VMware, Unix and Windows. She holds a PhD in Engineering from Stanford and has several publications from her research work as well as a patent at Google on Android IoT protocols.
Join us for KubeCon + CloudNativeCon in Barcelona May 20 - 23, Shanghai June 24 - 26, and San Diego November 18 - 21! Learn more at https://kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy and all of the other CNCF-hosted projects.

A

Thank you, I'm a pronoun Sena product manager for kubernetes at Google. You know, prior to Google, I worked on a software monolith. It kind of looks like the cinnamon donut actually, but this was at an enterprise company and it wasn't as large as the or ulu. This is the largest monolith in the world in Australia, but we did have a lot of technical debts and we were in the process of rewriting our code to be more modular and we nearly risked our business doing that.

A

So when I came to Google, it was like coming to another universe with Borg developers, didn't worry about adding machines or doing upgrades, and the operations team made a lot of improvements and innovations. In the background, the constraints of the Borg infrastructure ensured that applications were scalable efficient and globally available. There were a couple of papers published by Google explaining how they did this, but it was still really pretty alien for the rest of the world. That is until there was a seed sown on github and over the last six releases.

A

That seed has really matured to. As you heard, nine hundred contributors 35,000 commits to be one of the most active projects and largely this is due to the community that evolved around it, and this community really refined the plug-in model for networking for storage, enabled deployments on OpenStack and on on-premise and other cloud providers and really built commercially supported offerings for real world deployment. So that's great, but there are still a set of engineering concerns that linger and I. Think many of you might have these questions just kubernetes skaila. Is it too heavy?

A

Is it hard to install? Is it only for stateless apps, so in true election style? Let's do a fact check against the progress on these? Can it scale up, so we define cluster scale using an API, latency and pod startup time metric.

A

It started with a hundred nodes, which is very humble, but in July of this year, kubernetes scaled up to 2,000 nodes per cluster and very soon after we had Pokemon, go really go global with their service using kubernetes on Google Cloud, and we are on track to demonstrate even higher node counts in the next couple of releases.

A

So maximum scale is great, but really auto scaling up and down at the pod level and also at the cluster and node level is really what enables the separation of concerns between app developers and operations. And lastly, Federation completes that scale-up story by allowing services to span multiple clusters. So I think this one is a check, but what about scaling down so is kubernetes too heavy for smaller sites. You know a couple of weeks ago, you might have seen that Walmart announced the use of kubernetes in many data centers at their logistics warehouses.

A

These are really small deployments and they're running kubernetes on traditional VMware and vSphere boxes, really very impressive. Okay. So if it scales up and scales down, is it too complex to use and install well, a lot of folks in the community are working on the day one install experience, cig lifecycle has introduced, cube ATM with just two command installation. K Ops has been making it simpler on AWS and Q cuddle run and compose make it easy to deploy or services. And, lastly, a mini cube makes the local installation experience for testing much easier day.

A

Two was arguably not that bad with deployments and upgrades and also auto scaling, but the dashboard UI has been significantly improved, as you saw in the last demo. Our Doc's are still a work in progress, so overall I would say this whole area is a work in progress. The next question is: when can I use kubernetes for stateful and batch workloads? There's really significant effort and progress in this area and users are doing this today. So this is a satellite picture of us.

A

Cornfields Descartes Labs uses jobs and machine learning in kubernetes to analyze these image and improve crop yield. Another example Comcast recently used stateful sets and contributed their work back for the yarn implementation.

A

Also native spark support is on the way. So I would say this area is a very fast work-in-progress. Stateful set is moving to beta and we are regularly adding support for stateful applications through curated home charts.

A

So sometimes people still question whether kubernetes is optimized for Google's cloud. While it's true that Google container engine deploys the latest release of kubernetes within days, it actually does so using pure upstream without any modifications- and you already saw earlier that the majority of contributions to kubernetes are come from outside of Google and in fact eleven of the talks at this conference are from users who are using it on premise or in hybrid deployments.

A

Google and the community continued to really invest in Federation and an extensibility to make sure that there's user choice, in fact sometimes there are projects that even we're not aware of that use. That extensibility did. You know that can scible is a project that enables you to use kubernetes to orchestrate non containers. Please look them up.

A

Lastly, I would say the most important piece is the managed service provider program that Dan Cohen announced today. That will really make support and partners available to enterprises everywhere, with that I think that we may even be ready for the world's largest monolith. Thank you.