Cloud Native Computing Foundation KubeCon + CloudNativeCon North America 2018 (Seattle), 15 Dec 2018

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Keynote: Open Source & Containers on AWS - Arun Gupta, Principal Technologist, Amazon Web Services

Description

Join us for Kubernetes Forums Seoul, Sydney, Bengaluru and Delhi - learn more at kubecon.io

Don't miss KubeCon + CloudNativeCon 2020 events in Amsterdam March 30 - April 2, Shanghai July 28-30 and Boston November 17-20! Learn more at kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects

Keynote: Open Source & Containers on AWS - Arun Gupta, Principal Technologist, Amazon Web Services

More users are running containers using open source technologies on AWS than anywhere else. We'll cover the state of open source for containers and applications on AWS, new product announcements for the container ecosystem, and our vision for the future.

To Learn More: https://sched.co/GsxD

A

I'm very excited to be here at Q Khan at Amazon. We innovate on behalf of our customers so that they don't have to do the undifferentiated heavy lifting so that they can focus on their applications which they enjoy. We are innovating at all layers of the stack, whether its containers, micro services or virtualization.

A

We know open source matters to our customers, and so it does to us. So when we innovate on behalf of our customers, we make sure that there is full of stream compatibility to that software. Let's take a look at it on how we are helping our customers here. In 2018, AWS customers launched 15 billion containers using Amazon, ECS, Amazon, eks and AWS or gate we announced Amazon e cares earlier this year, which is our managed kubernetes service. This was g8 in June, so we've been busy getting the service out and rolling it out.

A

The e Care Service is growing at an incredibly fast pace. Our customers are loving it it's available in four regions and we are constantly listening to customers on rolling it out to other regions as well.

A

Now, upstream, compatibility is of core tenet of Amazon. He cares as well what that means is whatever you're using in your on-prem data, centers or anywhere else. That's the same experience you wanna bring to AWS cloud, but have all the ill at ease that you are aware of from Amazon scalability reliability, availability at amazon offers. This graph shows another trend on how we have been contributing to upstream kubernetes. We're very excited about the trend you know since june.

A

After we reading the service, the upstream contributions have grown, and we only expect this to grow from here.

A

These are some of the projects that we have been involved with the first one particularly is providing an application load balancer to a kubernetes ingress resource. We are also involved with providing a container storage interface to Amazon EBS into the elastic block store. The key part to here is these are the projects that are we are involved with, but the first theme are already available as alpha features in kubernetes 113, we're very excited about it, and this model of operating more and more in the open source is only gonna grow and continue from here.

A

Let's take a look at what we are doing in the community. As you realize this morning, HPD was contributed to CN CF. We're very excited that one of the lead, maintainer zuv HCD, works at that Amazon and we're very excited about that.

A

I wanna give a shout out to weave works and Stella gent, who have built open-source tooling around a magani cask, which really simplifies on how you can easily create Amazon ETS cluster or as a developer, you can easily create a end-to-end deployment pipeline using Stella jammu AWS operator again provides you a first-class experience of provisioning, AWS resources using cube kernel or you've control, as you like it going forward. Our focus is gonna, be to take over testing of kubernetes upstream for AWS, and that includes your conformance scalability.

A

We're also looking at Windows worker nodes compliance and how that would work on AWS lost, cube. Con Adrian mentioned that we would like you to build your open-source career at Amazon, I'm, very exciting. On growing and building my open-source career at Amazon within amazon eks team, we are very actively hiring. We have a recruiter on the booth, encourage you to come and talk to us and build your open-source kubernetes career at amazon.

A

Let's take a look at micro services. Why do we? Why do developers or organizations do micro services? Well, it provides a consistent communication mechanism between teams and software and everybody else. It also gives you deep visibility into your application on how the application is going to work because of the scope of the micro service. It also allows you to recover from a failure beside an act and move on and last, but not the least because it to iterate faster.

A

You can have each micro service independently deployed upgrades scale, and that gives you a jollity for building those applications.

A

Well, one of the common recommendation is to use service match for observability and in order to use that service match, use a sidecar proxy essentially well. What's the reason for that? Well, it'd be couples your operational control from your application logic as application developer. You are more interested in building your application logic and let the sidecar essentially take care of monitoring, routing service discovery and concepts like that. It also avoids framework explosion.

A

If you're building your micro services across multiple languages- probably you are, then you have to maintain each language dependent framework to do all of those services with sidecar. You delegate that approach to the proxy and you continue focusing on your application and last, but not the least because you are using the sidecar. It gives you traffic control, it gives you simplified debugging, troubleshooting a consistent experience across all of your micro services.

A

Overall, it simplifies your process for migrating from a monolith to a micro service is very easy now, and why is definitely a common proxy that is used for service smash and I would like to congratulate on on why becoming the third graduated project essentially joining the league, the kubernetes and prometheus?

A

Now we support easty on eks, and that works very well, our customer music, but our customers really wanted to build something that would run not just on kubernetes across clusters, across container orchestration and even virtual machines. They wanted to see how this would work in an Amazon way reliable, scalable highly available.

A

They don't want their cluster to be using the resources for running the control plane for on why they they really want their cluster to use for their applications and last but not the least, you know, the important part is having an upstream compatibility of it.

A

They want to be able to to provide envoy for them to use, but yet have the ability to say we can use any envoy that is available in upstream, so we introduce WS half-match AWS, app mash at reinvent, which is basically a highly available multi-channel control plane, which is running as a managed service, and that would work across all of these features. The cool part here is the roadmap for it AWS app mash in out on github.

A

You can take a look at it and any changes that we are making to envoy are going to be upstream. That gives you the ability to bring your own envoy or use the Envoy that we are providing.

A

Finally, let's take a look at it. What we are doing now in the virtualization space? Well until now, you really have to pick sort of the fast start up times and iterate faster from the container side of it or you could use a hardware based virtualization security and work like oscillation of virtual machines with firecracker is not an either/or choice. You don't have to pick one or the other pi cracker is an open source. Virtualization technology that is purposely built for server less computing, whether it is functions or is containers it gives you a secure.

A

Multi can end lightweight micro VM, which is very well-suited for server less computing. Park record was built at Amazon, and this is what we use to power. It abuse, lambda and AWS, far gate Fergie, a fire cracker, provides you security from the ground up, it is using KVM based virtualization and it is a VM M or a virtual machine manager that sits on top of it because of a low memory footprint model. It really reduces the attack surface area. For you, it is really built for speed by design.

A

You can easily fire up a micro BM in less than 125 milliseconds, and that also gives you the ability to be able to launch 150 micrograms per second per host. If you have more number of holes, you can easily fire up that very micro, B, apps and last, but not the least because of a low memory footprint. It really allows you to run a lots of those micro VMs on a particular host. So, essentially, you can run thousands of micro beams on a host really in a highly multi-tenant way with firecracker.

A

You can see that we are investing the same deep investments into serverless computing as we did for in-situ. You can run firecracker on AWS bare-metal instances on any bare metal server or even nested virtualization, if you realize well, that was shown recently in a TDI cake series. That's an exciting part of it. With about six thousand stars in little over two weeks, firecracker repo has been on fire and there's a lot of exciting developments happening yesterday. We had a really good meeting with Carter containers, folks and really a deep engineering discussions are happening right now.

A

The team is actually building having those discussions with continuity, folks and that work is actually happening in completely in the open source and the continuity would allow you to run firecracker beneath kubernetes, that's a time.

A

So what I want to leave here is how we are helping you innovate for your customers. We are giving you a fully upstream kubernetes at scale, we're giving you an Y based service mesh control plane not just for kubernetes, but for a wide range of compute instances on EWS we're also giving you micro biomes for server less computing, and the most important part that I'm excited about is how we are ramping up our contribution to open source well. Thank you.