Cloud Native Computing Foundation Kubernetes on Edge Day EU 2021, 14 May 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Kubernetes at the Edge for Telcos - Abhinivesh Jain, Wipro

Description

Don’t miss out! Join us at our upcoming event: KubeCon + CloudNativeCon North America 2021 in Los Angeles, CA from October 12-15. 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.

Kubernetes at the Edge for Telcos - Abhinivesh Jain, Wipro

5G rollout is gaining momentum across the world and in next few years, 5G and Multi access edge computing will become part of our life. Considering the cloud native nature of 5G architecture and huge scale, Kubernetes becomes the defacto standard for Edge application deployment for Telco Edge. This session talks about Telco relevance of various kubernetes distributions, extensions and features. It also covers pros, cons, current state of adoption and future outlook for these. This session will be perfect start for you if you are starting your k8s journey in Telco world.

A

Hello all welcome to this session. Today, I'm going to cover kubernetes at the edge intel course. My name is binivej chen, I'm 5gh architect, working in bipro.

A

Some more details about me: I'm distinguished member of technical staff, I'm open source contributor, I'm an author, blogger and frequent speaker here- is the agenda for today, I'm going to start with the edge computing, and uh so this part is basically uh to give a brief overview of edge computing for people who are not aware of it and then followed by that. I will cover telco edge overview and then why kubernetes is suitable for telco edge.

A

And then I will list down all the requirements that telco have from each platform and then we will do a mapping of those requirements with kubernetes and kubernetes ecosystem products right to see the fitment.

A

And then I will show one telco reference architecture and towards the end, we will see on the cncf focus item around this telcos and kubernetes, and then we will conclude this okay. So let's get started.

A

So what is such computing as computing is all about bringing cloud capabilities to the edge. So it's like you, bring your compute network storage uh at the edge right rather than running things from your data center or from your cloud, which is again centralized kind of thing. Now edge computing is not a replacement for cloud computing. It is rather going to be a complementary right and edge. Computing will be useful for low latency uh requirement, related application and application, which require real time uh processing.

A

Okay, there could be several possible use cases and in the beginning uh we expect that the cloud gaming and high quality video delivery will be the early adopters, okay from edge computing market perspective, as per lfh state of the edge report. I have given the link here in this slide, so you can see the detail.

A

uh This global edge computing infra market will be 800 billion by 2028 right. So there is huge potential in this and lot of work happening in this area.

A

Okay, on the right side of this screen, you see uh there is a prediction from gartner that was done in 2018, where gartner said that 75 of enterprise generated data will be created and processed at the edge right by 2025.

A

Now, what is telco edge so in telco world edge computing is commonly referred as mobile, edge, computing or multi-access edge computing, which is new name of mobile edge community, and there is a standards body called etsy european telecommunications standard institute, so etsy has defined msc reference architecture and also they have come up with several mec apis.

A

Okay, so you can have a look at etsy reference architecture and other details. Now this telco edge is also known as service provider edge right here.

A

Edge location is within the telco network, so telco is primarily edge, which is within control of telcos right and it could be near edge or far edge. So when we say near edge, it is near to the core data center of telco right and far edge is opposite of it, so it could be as far as customer premise itself and this edge computing or the nec is related to 5g.

A

In fact, msc is going to be an absolutely necessary component for achieving 5g capability around your llc, which is ultra reliable, low, latency communication right. So 5g has three main capabilities: one is urlc, other one is embbb enhanced mobile broadband and third one is mmtc massive machine type communication.

A

Okay, so why kubernetes is suitable for telco edge number one scalability right so uh since edge deployment will not be like eight or ten, uh the typical data center deployment, they will be in thousands.

A

So uh we need a solution which is scalable right, where we can do deployment in thousands and thousands of sites right and which can scale according to the needs. So let's say if one particular 5g core component needs to be scaled up, it should be easily possible right. Then second, suitability parameter is agility, so kubernetes supports various application types right, so you can run let's say, aml workload using queue flow you can read, serverless using k, native and so on.

A

Now kubernetes is a container orchestration system, so it has very good orchestration and life cycle management capabilities and the key feature of kubernetes is it has portability across hybrid multi-edge cloud, so this feature helps in doing workload. Migration from.

A

Which could be in public cloud or which could be in another edge location, on premises, location, okay and then kubernetes also give a consistent user experience. It provides site resiliency by using the portability, then it is also aligned to 5g core component right. Let's say upf can be placed closer to the consumer right and other components. Also and kubernetes is also aligned to open ram and cloud native nature of 5g right. So there were major difference between 4g and 5g.

A

That 5g goes 5g rather is cloud is following cloud native architecture and kubernetes is a synonym to cloud native architecture. So if something is compatible with kubernetes, then it is assumed to be compatible with cloud native.

A

Now, let's see what are the three requirements of telco and how does kubernetes and its ecosystem map to these requirements, so uh various telco companies have so many requirements and I've classified those requirement into four areas. One is application area right, so from application perspective, what I mean to say is that whatever is the edge platform, that platform should be capable of hosting various types of application.

A

It could be ai application, iot application, serverless application, and there should be some match application. Lifecycle management option also right from network function hosting perspective.

A

Telcos are looking for platform that helps in hosting cnfs, which is cloud native network functions and vnfs, also, which is virtual network function right, which are running on vms right so and for private network. uh They are looking for vnf or cnn, based hosting right and since telco workloads are very network intensive. There are very specific features which are required like huge base, support or sri uv support.

A

Even multi interface support. Real-Time kernel support, cpu, pinning pneuma support uh ptp support. There are many more actually right. So these are the technical features which are required in the platform and from use case perspective, use cases could be related to cloud gaming related to video analytics related to iot edge and so on.

A

So now, let's see so how these things are mapped. So many things are supported, as is let's say, huge, face support as a user. This in kubernetes and few things we can uh implement using related products. Projects right, like cube flow, can be used for implementing either edge or k native can be used for serverless.

A

Similarly, cuboid can be used to be in a hosting right and for tiny edge or iot edge. We can use kubernetes distributions, like k3s cube edge micro creators. These are very small uh small footprint based distributions.

A

Okay, so moving to reference architecture, so I mean from previous slides. It is clear that we, we need kubernetes to run uh telco edge kind of workload, but now how I mean what is the reference architecture? How do we structure these components right? So this is one of the sample reference architecture that I have taken from acrino, where I'm also one of the active member. So uh in acronym uh kni blueprint, which is kubernetes native infra blueprint, the architecture is that at the bottom of it you have a bare metal server.

A

You have virtual machines, you have a public cloud based uh machines like uh aws cc2 instances right and on top of it, you are running either uh operating system like core os, which is very suitable for container related workload, or you might be running some real time kernel like centos, rt right and then on top of it.

A

You are running some software defined storage like chef, and you have a multi network interface supported option like multis or cryo for container runtime, and then you run okd right, which is open source kubernetes distribution, which is open, shift, open source version right, and uh on top of uh this kubernetes, you run kubert during various virtual machine based workload, or you run container-based workload directly as mec apps, and only so. This is one of the reference architecture.

A

You can also refer to mec reference architecture given by etsy right. uh That very nicely explains how these components should be assigned. What is the role of each of those components and so on?

A

So it's a recommended architecture and many companies follow that in fact, right and now this is one of the reference architecture and there could be several different possible reference architecture right nowadays. Various vendors have also started giving their as a reference architecture like aws has given one and let's say: vmware has given one and so on right.

A

So, but more or less components will be like this only right.

A

Okay, now, let's see what what is the focus of cncf or what cncf is doing in this area right so cncf objective is that they want to do best utilization of kubernetes features by telco operators and these developers, who are working on cloud native network function.

A

So for that there is a telecom user group. I have given the link also, if you are interested to join that group or want to get more detail. So there is a telecom user group which is doing several things, including they have cnf test pad.

A

Also- and there is a cnf working group networking spatial interest group right, so a lot of activities are happening and cnf testbed is a very good initiative where one can test their cnf on kubernetes and they also have openstack based environment, where they can compare that with corresponding venus right and do the benchmarking in terms of performance and in terms of other parameters.

A

Now there are a few projects like network service mess right, which is running to handle this kind of requirement of telcos. Then there is cuberd project I have already covered about uh kubert. Then there is linux foundation edge project right. This lfh is a dedicated project for edge computing, and there are so many edge stack uh related to telcos.

A

So you can have a look and actually know h. Stack is one of the project of lfh out of six or seven projects, and acronym has at least 20 different blueprints for different electric use cases so yeah. There is a huge amount of work happening now in terms of current challenges, so is kubernetes all geared up and ready for dengue, uh probably not not hundred percent. So there are some of the challenges which are listed here.

A

Number one bnf hosting on kubernetes is still a challenge, because uh cubert is uh getting matured at very fast pace. Probably it will be ready in next few months also, but right now it is not ready for vnf hosting for getting telco great performance right. So that is one area. Second is multi-cluster management, so, as you know that uh edge deployment will be running in thousands, so we need a multi-cluster management solution right from where we can have a single pane of glass of everything that is happening.

A

uh In my let's say, hybrid multi-edge cloud right: there are several open source projects which are trying to cover this flow, and there are vendor-specific uh solutions like uh aws has eks anywhere or google has anthos and then azure arc is another solution. So yeah I mean vendors are uh coming up with such solution. I think reddit also has advanced cluster manager, acm solution, okay, so uh next challenge is automated kubernetes cluster provision.

A

So this is very much required because when you are dealing with a large scale, you need one click deployment of not only kubernetes but uh all the things which are required alongside kubernetes right. So you need, let's say uh various operator to be deployed automatically which are required. That's a chef, rook operator or something else so that, after that, your cluster is up and running without any manual intervention, so that is required.

A

Then uh another feature which is uh required is kubernetes cluster label. Dr now, uh what I mean by cluster level, dr is when let's say I am having one kubernetes cluster at h: type, one and another cluster at h. Two and my s site goes down. How do I move all the workload from side one to se, two right, so that kind of dr option is required.

A

uh Kubernetes as such is not built with that kind of architecture in mind right. It is always single cluster kind of architecture, but we can achieve this kind of.

A

Functionality by using various storage solution, like with sync replication, async replication at the storage layer, but those are quite challenging to implement right so, and migration in such cases is not very straightforward and easy. Even some of the best storage uh container native storage vendor are still uh figuring out the way to provide uh seamless migration from one trustworthy cluster right, okay, so the last two challenges one is enough to cnf. Conversion is quite challenging, and deployment complexity is also very high.

A

The key takeaways here telcos are becoming software defined earlier. They were very much applied, centric and all, and not only telcos, but even the oems are also becoming softer different and telcos already getting ready for playing bigger role right and cloud kubernetes open source. These are going to be essential, technological components in telco's journey right, educative applications are also being built and h, native application will be built on top of kubernetes, and uh so so far we were talking about edge edge and kubernetes, but kubernetes role is not limited to edge.

A

In fact, kubernetes is also used by leading 5g core vendors.

A

And they are using it for 5g, core and various rand. Vendors are also using kubernetes for providing uh open ram and so on right, so kubernetes is coming in big way in telco landscape. So that's a message here so with that we have reached a question session. So if you have any question you can type it in the chat box and I'll answer and if not, then we have reached towards the end of the session. So thank you very much for attending this.

A