Cloud Native Computing Foundation Cloud Native + Open Source Virtual Summit China 2020, 11 Sep 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Shaping the Future of Cloud Native Infrastructure at the Age of AI, 5G and Edge - Ying Xiong

Description

Don’t miss out! Join us at our upcoming events: EnvoyCon Virtual on October 15 and KubeCon + CloudNativeCon North America 2020 Virtual from November 17-20. Learn more at https://kubecon.io The conferences feature presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects.

Shaping the Future of Cloud Native Infrastructure at the Age of AI, 5G and Edge - Ying Xiong, Futurewei

The combination of next generation AI and 5G applications poses major challenges for cloud infrastructure platform to support low latency and high scalability requirements of future cloud workloads. The convergence of Edge with the Cloud is driving the innovation of cloud platform architecture for unified management and large scale geographic deployment across data center and edge infrastructure. In this talk, we introduce a new open source project that provides a highly scalable and truly multi-tenant cloud native infrastructure to build next generation public and/or private cloud. We will present two open source components - Arktos, the large scale and partitioned compute cluster with unified management of cloud resource and workload including VMs and container, and Mizar, the high performance and low latency cloud networking solution that leverages advanced network technologies such as XDP/eBPF and Geneva protocol. Together, Arktos and Mizar deliver a new generation of cloud platform aiming to meet the challenges in the age of AI, 5G and Edge. In this talk, we will also give the use case of autonomous driving that can benefit from the new open source project.

https://sched.co/dINS

A

Hello: everyone welcome to this lightning talk: shipping, the future of cloud native infrastructure in the age of ai, 5g and edge. My name is I'm with virtual technologies. Today, in this talk, I'm really excited to introduce you a new open source project that is designed for next generation applications.

A

So let's go ahead and get started first, let's look at some of the challenges that we are facing today. As the cloud adoption grows explicitly, the infrastructure platform needs to manage larger and a larger scale of compute and network resources across your graphic area.

A

The customer always asks for faster and faster provisioning of the resources or the applications.

A

Additionally, there is an increasing need for cloud operators to unify the management of diverse type of resources such as biometers, vms containers and serverless functions to just simply reduce the management cost and simplify overall operations.

A

With the emergence of the new type of workload such as ai 5g and iot apps, we see a train that traditional cloud computing and edge computing are being converged into one distributed cloud.

A

This train requires that the current infrastructure technologies need to be redesigned or optimized to support applications running at resource constraint, htmls that are closer to the customer and the customer data.

A

As a result for those new type of work notes, we need a cloud instructor to be able to internalize place where the application should run in order to improve custom experience with low latency on a high circle.

A

Last but not least, the train in the era of qualitative continuous due to its extremely scalable and extensible design. We believe the modern cloud infrastructure should incorporate this qualitative design pattern into the platform to meet the challenges facing by new generation of applications.

A

So now I'm pleased to announce and introduce you, the new open source projects called centers.

A

The goal of projects and tariffs is to provide a new infrastructure platform in a more collaborative way for any company to build scalable and unified infrastructure for the public or private cloud. Centerra's platform is designed from its beginning with hyper scale, edge computing and fast compute network provisioning in mind.

A

Centaurus projects actually consist of two side: projects will name them arctus and the misa more detail on actors and the misa will be discussed in the following slides.

A

This is overall architecture of projects and terrors. I won't be able to deep dive into this architecture in detail, but I would like to mention a few highlights.

A

First, regional scalability: one of the reasons for the projects and terrors is able to support a cost of 300 000 computer nodes and be able to provision and managing more than 10 million network endpoints within a region. So every component in this architecture can be partitioned to handle large number of requests.

A

Second, this architecture natively supports edge computing, is built in to schedule, run and manage applications at age size.

A

Lastly, this architecture also offers the same api experience for both the vm and the containers, as well as for other type of workloads. He provides application-centric api is a platform for both resource and application. Orchestrations.

A

Centara's output components is derived from well-known kubernetes architecture. However, we made some major redesigns and extensions to the core of kubernetes.

A

Here is a list of the changes that were made partitioning and skin scalability in architectures, api server, etc. Store controllers, schedulers all can be partitioned for large scale deployment.

A

Multi-Tenancy architectures introduce the true and hard model tendency into the kubernetes and make the platform a real modern data platform unified the runtime actors unify vm and the container launcher into one so that the platform manages vm in the same way as he manages containers.

A

Architects introduce new network object types in the corresponding operators. This is a part of multi-tenancy model so that containers or vms are created within subnet or vpc, and they are naturally isolated.

A

Built-In edge computing actors out of the box can manage computer nodes at 80 sites and orchestrate applications running at a resource constraint, age.

A

Similarly, centara's mesa component is a network component that provisions and manage the virtual network and the process network traffic. Visa itself has two components: data plan and the managing plan. The data plan is built on top of a high performance. Xdp, slash evpf programs, xdp programs running within the os corner to perform routing, switching and another balance of network traffic.

A

These are using natural partitioning of a cloud network such as vpc and subnet to scale network endpoint, provisioning for attending vms or containers, as well as for fast networking. Forwarding visa management plan runs within a kubernetes and using kubernetes crd and operator framework to manage virtual networks and the network package routing policies.

A

That's all for me for just a short talk. Thank you very much for your listening. Please meet our virtualbooth for more detail about projects and terrors. Thank you very much.

B

A

A

A

A

uh Dear government is uh any plan to host this two open source projects: actors on the visa in a window, neutral organization, um uh women's.