5G Network Slicing

Session Abstract:

The "one-size-fits-all" network paradigm employed in legacy mobile networks are no longer suited to efficiently address applications like machine-type communication, ultra-low latency communications and enhanced mobile broadband content delivery. In comes 5G network slicing, a network architecture that enables the multiplexing of virtualized and independent logical networks on the same physical network infrastructure. Here to shed some light on 5G Network Slicing are, Juan Carlos Garcia Lopez, SVP of Technology Innovation and Ecosystems at Telefonica, next is Robin.io’s Mehran Hadipour, VP of Business Development & Tech Alliances and on the end is Pallav Paliwal, Chief Architect of 5G/ Cloud at IBM.

Executive Speakers:

Juan Carlos Garcia Lopez - SVP Technology Innovation and Ecosystems, Telefonica
Mehran Hadipour - VP Business Development and Tech Alliances, Robin.io
Pallav Paliwal - Chief Architect | Network Cloud Leader, IBM

Transcript

5G NETWORK SLICING

Abe Nejad: The one size fits all network paradigm employed in legacy mobile networks are no longer suited to efficiently address applications like machine type communications, ultra-low latency communications, and enhanced mobile broadband content delivery. In comes 5G network slicing, a network architecture that enables the multiplexing of virtualized and independent logical networks on the same physical network infrastructure. Here to shed some light on 5G network slicing are Juan Carlos Garcia Lopez he's senior vice president of technology innovation and ecosystems at Telefonica.

Next to him is Robin’s Mehran Hadipour, vice president of Tech Alliances. And on the end as Pallav Paliwal, he's chief architect of 5G and cloud, at IBM.

Gentlemen, welcome.

Thanks for being here. Of course, we're at day two here at the event in Barcelona. It's good to see everybody in person by the way, and not just virtually. Pallav, we haven't done this, but it's good to have you, and again, thanks for your time. Mehran of course we're in the Robin booth here at the show and it's good to have Mehran from Robin.io to participate as well and as always, first of all, it's good to have an operator’s perspective on all of our sessions, but particularly with Juan Carlos, who's always generous with his time with us.

Juan: Welcome to Barcelona.

Abe Nejad: I appreciate it. Thank you very much. Okay, let me start with you Juan Carlos, how is Telefonica approaching 5G network slicing?

Juan Carlos: Well, we have been trialing the last two years a number of use cases on 5g. So you can see our webpage more than 50 cases and also in the booth, in our [01:39 inaudible] booth more than 90 and most of these 5g used cases are somehow using some kind of slicing. It is of course. Semi manual slicing. I would say is creating a specific [01:52 inaudible] for certain applications of customers like public safety services, video surveillance. And some other applications. So it's not that fully fledged end-to-end slicing.

For this, we will require to have the capabilities to also allocate dynamic resources on the radio, on the transport and also on the edge. And this is something that will come over time. We are working actively in the different standardization bodies to make it happened. So IETF for the transport for instance, Oran and [02:28 inaudible] for the radio and also bodies like Telefonica [02:34 inaudible] or [02:34 inaudible], where we are defining the mechanisms for orchestrating all these different domains for the [02:41 inaudible] slicing.

We will very likely see first, not this full end to end. We will see first capabilities being offered as a service that will allow the customer to control certain parameters of the connectivity. And in some cases create a [03:01 inaudible] slice for certain application. And this will be probably the beginning and the start of all this. We have launched. And we announced yesterday, the launch of an open source project in the [03:14 inaudible] foundation with nine other operators, with the cloud providers, with [03:20 inaudible] providers, with intention to define, standardized this APIs. Not to expose the network capabilities to third parties. And this is our first step towards this exposure of more complex capabilities like networks slicing. That is something that we will see probably in 2, 3, 4 years from now.

Ade Nejad: Pallav. Let's talk about network as a service. Juan Carlos alluded to that. Let's talk about how 5G networks or just network slicing as part of that service portfolio.

Pallav: So if you see network as a service, is itself a cloud model. And you mean to say that disaggregated clouds coming over as a service and you have a flexibility to basically define your own networks, customize those sources. And the same concept is coming out into the slicing, because it's a highly given slicing, which you basically say that the slice you create, you customize, you customize in, RAN, you customize in transport, you customize basically into the overall transport layer and you need to decide, what exactly you need to customize and put it and the network as a slicing service there.

Abe Nejad: Mehran anything to add?

Mehran: Sure. I think the interesting aspects that kind of complicate them at the same time enables this, is as you get towards the diversification and [04:34 inaudible] of telco that deploying Kubernetes as a platform or delivering network functions as a service. If you had a Kubernetes infrastructure that was multi-tenant and could be configured automatically from a central location across all the elements of the network function from ran to call to the rest of the infrastructure that automation could be used to kind hyper accelerate, how we do slicing. Slicing could be done across the infrastructure and the software stack. And a lot of that could be automated using Kubernetes and Kubernetes constructs. And that's what Robin is trying to do.

Abe Nejad: Juan Carlos, you mentioned you have a number of used cases in your booth right here at this event. And I went by there actually this morning. And when you say 90, there really are 90 used cases. I believe it, when I walk through there. So give us sort of into the looking glass of maybe a couple of those used cases, maybe something that we haven't heard before because we talk about augmented reality, low latency, high latency applications like IOT versus AR. Is there something else, maybe an example that we haven't heard?

Juan: Well, I mean the latest ones that we have even launched as a service to customers, this service for drone surveillance of critical infrastructures, like power grids or railway and railway lines and all this is done through, orchestration on drones that are making this inspection automatically. This is one of the examples. We have also for instance, a case of remote video production, where we are able to avoid the need of a content provider to have a track close to the event, to receive all the signal and process the video.

We can do this now from the edge. And using a specific slice that provides the [06:26 inaudible] bandwidth that is required to take all this raw video to the processing center at the edge. So these are examples that we have for many other.

Abe Nejad: Yeah. IBM booth, same, a lot of activity there. Again here at the event in the network slicing space, can you give us again a use case or two where you're able to leverage for low latency applications and also high latency applications.

Pallav: So if you see the ultra low latency and ultra reliable communications, there are three, four important things. The concerns basically which has to be addressed from the networks slicing scenario. You take up example of AR and VR, what exactly you need, the first thing which basically you need, is how you go to the multilink connections there. And that comes from the different network functions, whether it's a 5g network port function, or maybe you talk about, how the IP address of the multi broadcasting needs to be done there.

And then the second point, which is very important is how you define actually those latencies, how you define those bandwidths and who will take basically that orchestration control, who will define those [07:29 inaudible] models into the closed loop, is the kind of analytics and then orchestration taking into place and then complete planning of these slice mechanisms, how you do that. So we have a product like cloud [07:39 inaudible] for network automation, which normally says that you can plan your slices. You can basically have a closed loop control. You can define the entire ran portfolio, the slice management from the core side and interlinking towards, through the transport. So what exactly you need, you put that into the slice and you model those basically profiles into the sub slice instances there.

Abe Nejad: Mehran here in the Robin.io booth, any use case or test case that you can?

Mehran: Quite a number of use cases that demand network slicing that we see in the market, especially around some of the industry [08:12 inaudible] used cases, video analytics, transportation. We have a port project, for example, with one of the operators that are trying to deploy automation layers. The challenge I see is that there are a lot of infrastructure challenges on the outbound side, what's being deployed at the far edge and orchestration of that with the rest of the infrastructure happen at the core is quite critical to have a slice management model that actually works and is automated and be able to be managed and resourced, allocated, and also build accordingly all that has to be part of the infrastructure automation layer.

And what we do with MD [08:57 inaudible] is do central automation, not only on the core components, running on Kubernetes, on the data center, but all the small elements like the single servers doing processing at far edge. So this allows you to have an end to end view of the slice and also manage it. If you don't have that automation in place, infrastructure doesn't matter. You have to make sure that you can do that centrally. You can do it easily and you can do it automatically.

Abe Nejad: So Juan Carlos, as we move from, let's say, in theory behind network slicing the test lab and now to these full scale deployments of 5g network slicing, why is it just as important as how you automate something as what you automate?

Juan Carlos: Well really, automation is probably not something new for us. So we were already implementing automation solutions for the last 20 years. The difference with networks slicing, anything that allocates resources in the network is that you need to make a very precise allocation just in the moment the customer needs it. And for the time the customer needs it. And this requires that you even are able to anticipate to the needs, to have the network prepared for that. So you need somehow to make the network predictive.

And this implies, the extensive use of analytics, of machine learning, artificial intelligence, to understand the behavior of the network of the customer and the services, and then try to make a smart allocation of the resources for network slicing. If you don't do that, what you will have is probably not very efficient and non-sustainable network slicing.

Abe Nejad: Yeah. Pallav, how you automate versus what you automate?

Pallav: So we talk about automation. Automation starts from day zero, day one and day two use cases. So day zero comes, how you plan your slice or how you plan basically your clouds. And then it comes to the execution of all those, whatever you plan actually into the cloud and application models. The third part, which is important is you need to pick up what you need to automate, whether you want to focus on the day two or maybe on the day, zero and day one. Day, zero and day one, in terms of the massive deployments of the 5g specific operators, or maybe day two, because day two activities like scale and scale out, backup management, log management integration towards to the [11:14 inaudible] that has to be automated. So day two is a common requirement but day zero and day one, in more terms, the scales, like suppose if you take up an example of Oran.

If I need to basically put 1000 of sites, I need a massive automation to do that.

And people should understand, that it's not about the complexity to bring from a service integration. It's about the innovation in the service integration, which these software’s can bring actually. So the person who is sitting actually and configuring the networks is not important that he should know open ramp, or he should basically know transport, or he should understand those complexities. He should be sitting into the portal, defining it. You know what? I actually need it from the system. He should define the portal, the catalog, the design studio, and the approach is very clear, in a zero touch, he should be having the outcome coming out from the systems there.

Abe Nejad: Mehran Robin.io’s perspective on automation?

Mehran: You know, we always feel that automation is probably not only the number one impediment of complex network service delivering efficiently. It could also provide some opportunities for enhancing the network on automatically. You asked the question, what would you automate? So what if you could probably automate the actual infrastructure based on the performance requirement and changing the frequency of the service to consume less power. You have a project just like that with Intel using AI tools and observability framework on top of Robin, that takes a look at what's being run on the network and updates the server configuration to minimize the power consumption where power is not needed.

You're adding another slice. You need more power, you can jack up the frequency. So there's a lot of things that people are not looking at automating that could really have significant benefits. So I think focusing on, okay, let's start taking a look at what should be automated and what could be automated. That's one advice. The second advice is, think of automation end to end. Automating a certain component is actually counter productive. So, you need to look at end to end automation from service orchestration, service delivery, and lifecycle operations has to be automated.

Abe Nejad: So Juan Carlos, I'm going to wrap with this question and I'll go down the line here. I was going to ask it in a certain way, but I'm going to change that now. So let's take your demonstrations here at the show for 5g network slicing or network slicing. What would that demonstration look like a year from now at the same show? What would be the difference?

Juan Carlos: Well, our expectation is that we have much more automation or much more facility in getting access to the facilities of the network. Today the integration of new solutions is costly in terms of effort and time. And the expectation is that we create over our networks, a layer of service APIs that will allow us to deliver services more rapidly, to make the network capabilities and community capabilities easier to use okay and faster to deploy and operate. And this is going to happen based on the Alliance we mentioned, we launched yesterday.

As I mentioned to you the camara project, together with the leaders foundation and the ESME. So working together to enable this set of capabilities on top of the network, to facilitate the service development and service deployment. And this is in one year from now, as you say, what we will see is that these cases will be much easier for us to deploy, to roll out and to be developed, both for Telefonica and for [14:52 inaudible].

Abe Nejad: Pallav same question 5g network slicing use cases, more deployment, less sort of test lab in theory. Where do you stand on that?

Pallav: So, okay. Let me start from there because the complexity in the slicing today is as what Juan Carlos mentioned, it's about the API management part. From the lab when you talk about the APIs and when you talk about how your service orchestration or automation engine talk to the different domains. So it's RAN code transport quite complex in nature. Then again, how the API management needs to be handled from top and down. And that API needs to be very standardized. And it needs a lot of efforts in the lab because once you test all those API management, all the standard interfaces will come down into the manner. And it needs to be understood. Once this standardization is being done, the automation is quite easy and from the lab you can push those things into the production directly there.

Abe Nejad: So Mehran still some work to do in 5g network slicing. Where do you see that area this time next year?

Mehran: Good question. I think slicing is almost the must have, especially in some of the new model of service delivery, like neutral host for example and has to be something that most operators have to deal with. The level of automation is improving daily. There's a lot of things where people are doing to try to improve the orchestration. IBM mentioned the cloud piece. We see that the infrastructure level of automation, for example, multi-tendency, Kubernetes distribution that can host multiple network functions independently and logically on the same network.

The fact that you can run VNF and CNF today on the same Kubernetes distribution. The fact that you can create a network function across a common hardware and deliver to different operators and have usage and consumption reporting, all that stuff is there. What is needed is really practical, deployment and experiences and people using those tools to really deliver real networks. And I see that happening as we speak. And I think next year, we will have to point to several live examples of where slicing is deployed, how it's being used, is going to be a fact of how network functions are delivered tomorrow.

Abe Nejad: Yeah. We've heard several sort of buzzword and keywords. Did you want to say something?

Juan Carlos: Well, yeah, just to point out, it is true. We need to exercise and start working with these orchestration tools. This is very relevant that we have standard APIs on these platforms. And the reason for this, this way, application developers will invest time in using them and even platform providers will also invest time in implementing these tools that are absolutely necessary as you say, for network slicing.

Mehran: Yeah. That makes a lot of sense. I think one of the things you have tried to do is to do a workflow model and create these workflows for slash creation and slash management and so on that have modular components. Some of these APIs are all not there, and that is going to get time to develop, but you can insert the modules for different network providers or network function providers. So you can have whether using somebody else's code tomorrow or somebody else's code right now, different ran and so on. You don't have to change your workflow. So the workflow stays the same, your automation layer stays the same. Now you have to add more APIs and [18:30 inaudible]. You can't just do it all at once.

Pallav: And it's important, from the workflow management, we need to understand, the automation engines should be able to understand how the intent of the lifecycle management will follow there, because there are a lot of workflows. There are a lot of sequential steps and tomorrow if you want upgrade your slice or upgrade your application, how the sequential activities are going to be impacted. So that's the intent and that's the behavioral testing, which it needs to be addressed because once you deploy a slice or before the cloud infrastructure is ready, how you will endure, that the platform is up to date to have the application on top of it, or any slice, those kind of behavioral testing needs to be done.

Abe Nejad: Yeah. I was going to say just a second ago that there's a few buzzwords going around this event as they do at previous events, Open RAN would be one, Metaverse Ready Networks, which we're going to have a session on right after this with Facebook, now called Meta. But the third one is 5G network slicing. And certainly something that there's no shortage of conversations around. We'll probably be talking about it far longer than today. So again, good to have the operator perspective as always, Mehran good to have you. It's your first time you and I are doing this. I'd love to do it again and Pallav thanks for your time. Appreciate it.

###

For any inquiries, please email anejad@thenetworkmediagroup.com

Abe NejadMarch 11, 20225G RAN