 Hello all, welcome to this session. Today I'm going to cover Kubernetes at the Edge in Telcos. My name is Viniwesh Jain. I'm 5G edge architect working in Vipro. Some more details about me. I'm a distinguished member of technical staff. I'm an open source contributor. I'm an author, blogger and frequent speaker. Here is the agenda for today. I'm going to start with edge computing. This part is basically to give a brief overview of edge computing for people who are not aware of it. Then followed by that I will cover telco edge overview and then why Kubernetes is suitable for telco edge. Then I will list down all the requirements that telco have from Edge platform. Then we will do a mapping of those requirements with Kubernetes and Kubernetes ecosystem products to see the fitment. Then I will show one telco reference architecture. Towards the end we will see the CNCF focus item around this, telcos and Kubernetes. Then we will conclude this. Let's get started. What is edge computing? Edge computing is all about bringing cloud capabilities to the edge. It's like you bring your compute network storage at the edge 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. Edge computing will be useful for low latency requirement related application and application which require real-time processing. There could be several possible use cases and in the beginning we expected the cloud gaming and high quality video delivery will be the early adopters. From edge computing market perspective as per LFH state of the edge report I have given the link here in the slide so you can see the detail. This global edge computing info market will be $800 billion by 2028. There is huge potential in this and a lot of work happening in this area. On the right side of this screen you see 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 by 2025. Now what is telco edge? So in telco world edge computing is commonly referred as mobile edge computing or multi-axis edge computing which is new name of mobile edge computing and there is a standard body called Etsy European Telecommunications Standard Institute. So Etsy has defined MEC reference architecture and also they have come up with several MEC APIs. 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. Edge location is within the telco network. So telco is primarily edge which is within control of telcos 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 and far edge is opposite of it. So it could be as far as customer permissible. And this edge computing or the MEC is related to 5G. In fact MEC is going to be an absolutely necessary component for achieving 5G capability around URLLC which is ultra reliable low latency communication. So 5G has three main capabilities. One is URLLC, other one is EMBB, advanced mobile broadband and third one is MMTC, massive machine type communication. Okay so why Kubernetes is suitable for telco edge? Number one is scalability right. So since edge deployment will not be like 8 or 10 the typical data center deployment there will be in thousands. So 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 AIML workload using Qflow, you can run serverless using Knative and so on. Now Kubernetes is a container orchestration system so it has very good orchestration and lifecycle 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 one edge site to other edge site right which could be in public cloud or which could be in another edge location on-premise edge location. Okay and then Kubernetes also give 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-run and cloud native nature of 5G right. So there were major difference between 4G and 5G that 5G goes 5G rather is cloud is following cloud native architecture. And Kubernetes is synonym to cloud native architecture. So if something is compatible with Kubernetes then it is assumed to be compatible with cloud native. Now let's see what are the three requirements of telco and how does Kubernetes and its ecosystem map to these requirements. So various telco companies have so many requirements and are classified those requirements for 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. It could be AI application, IoT application, serverless application and there should be some major application lifecycle management option also right. From network function hosting perspective 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 they are looking for VNF or CNF based hosting right. And since telco workloads are very network intensive there are very specific features which are required like huge pay support or SRIOV support, even multi-interface support, real-time kernel support, CPU pinning, NUMA support, 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. So now let's see so how these things are mapped. So many things are supported as is let's say huge pay support this is already there in Kubernetes and few things we can implement using related products projects right. Like Qflow can be used for implementing AI at the edge or Knative can be used for serverless. Similarly Qboard can be used for VNF hosting right. And for tiny edge or IoT edge we can use Kubernetes distributions like K3S, QBedge, MicroK8, these are very small footprint based distributions. Okay so moving to reference architecture so I mean from previous slides it is clear that we we need Kubernetes to run telco edge kind of workload but now how I mean what is a reference architecture how do we structure these components right. So this is one of the sample reference architecture that I have taken from Ekreno where I'm also one of the active member. So in Ekreno KNI blueprint which is Kubernetes native infra blueprint the architecture is that at the bottom of it you have bare metal server you have virtual machines you have public cloud based machines like AWS EC2 instances right. And on top of it you are running either operating system like CoreOS which is very suitable for container relative workload or you might be running some real-time kernel like CentOS RT right. And then on top of it you are running some software defined storage like Chef and you have 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 OpenShift open source version right. And on top of this Kubernetes you run Kupert you run various virtual machine based workload or you run container based workload directly as MEC apps and all. So this is one of the reference architecture you can also refer to MEC reference architecture given by Etsy right. That very nicely explains how these components should be assigned what is the role of each of those components and so on. 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 edge reference architecture like AWS has given one and let's say VMware has given one and so on right. So but more or less components will be like this only right. Okay now let's see 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 right. 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 CNCF testbed also and there is a CNCF working group networking spatial interest group right. So a lot of activities are happening and CNCF testbed is a very good initiative where one can test their CNCF on Kubernetes and they also have open stack based environment where they can compare that with corresponding VNF right and do the benchmarking in terms of performance and in terms of other parameters. Now there are a few projects like network service mess right which is running to handle this kind of requirement of telcos and there is Cupert project I have already covered about Cupert. Then there is Linux foundation edge project right this LF edge is a dedicated project for edge computing and there are so many edge stack related to telcos. So you can have a look and actually know edge stack is one of the project of LF edge out of six or seven projects and actually know has at least 20 different blueprints for different 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 telco edge? Probably not, not 100 percent. So there are some of the challenges which are listed here. Number one, VNF hosting on Kubernetes is still a challenge because Cupert is getting matured at very fast pace probably it will be ready in the next few months also but right now it is not ready for VNF hosting for getting telco grade performance right so that is one area. Second is multi cluster management so as you know that edge deployment will be running in thousand so we need a multi cluster management solution right from where we can have a single pane of glass of everything that is happening in my let's say hybrid multi edge cloud right. There are several open source projects which are trying to cover this scope and there are vendor specific solutions like AWS has EKS anywhere or Google has Anthos and then Azure Arc is another solution so yeah I mean vendors are coming up with such solution I think Red Hat also has advanced cluster manager ACM solution okay so next challenge is automated Kubernetes cluster provisioning 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 all the things which are required alongside Kubernetes right so you need let's say a various operator to be deployed automatically which are required that's a chef root operator or something else right so that after that your cluster is up and running without any manual intervention so that is required. Then another feature which is required is Kubernetes cluster level DR now what I mean by cluster level DR is when let's say I'm having one Kubernetes cluster at side one another cluster at side two and my side goes down how do I move all the workload from at side one to side two right so that kind of DR option is required 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 functionality by using various storage solution like the sync replication and sync replication at storage layer but those are quite challenging to implement right so and migration in such cases is not very straightforward and even some of the best storage container native storage vendor are still figuring out the way to provide seamless migration from one cluster to another cluster right okay so the last two challenges one is vnf2 cnf conversion is quite challenging and deployment complexity is also very high the key takeaways here telcos are becoming software defined earlier they were very much appliance centric and not only telcos but even the VMs are also becoming software defined and telcos are ready getting ready for playing bigger role right and cloud kubernetes open source these are going to be essential technology equal components in telcos journey right h native applications are also being built and h native application will be built on top of kubernetes and so so far we were talking about h and kubernetes but kubernetes role is not limited to h in fact kubernetes is also used by leading 5g core vendors and they are using it for 5g core and various ran vendors are also using kubernetes for providing open ran 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 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 to a stand of the session so thank you very much for attending this thank you