To embed our video on your website copy and paste the code below:
<iframe src="https://www.youtube.com/embed/YrwZ1M-rvqE?modestbranding=1&rel=0" width="970" height="546" frameborder="0" scrolling="auto" allowfullscreen></iframe>
Guy Daniels, TelecomTV (00:12):
Hello, you are watching the Open Ran Summit part of our year-Round DSP Leaders Coverage. I'm Guy Daniels. Now as the development of Open Run continues and with commercial deployments now underway, telcos are investigating how to use AI to automate and optimize their radio access networks. Joining me now to explain more is Paul Miller, chief Technology Officer for Wind River. Hello, Paul. It's good to see you again. Let's start off by looking at the challenges that CSPs face. What are the biggest challenges as they roll out and start to manage their open RAN networks?
Paul Miller, Wind River (00:53):
Yeah, thanks guys. So it's really interesting what's going on now. We've reached a point in the evolution of open ran and virtual ran deployments where we now have several years of deployment at very, very high scale happening in multiple geographies and it's really exciting to see that because we're obviously seeing the fruition of many years of investment for many companies in that ecosystem bringing open ran to reality. And so the feasibility and performance are now proven to be superior than legacy ran, but there are new challenges emerging as we've gotten to this stage of maturity in deployment. Obviously it's one thing to build an infrastructure and an application that functions, it's quite another thing in a telco service provider to build this application so it can be maintained and operated at full scale in a telco service provider network. As we know, many 5G deployments have over a hundred thousand sites that are used for the VDU in the case of the open ran. And to manage that as a fully distributed architecture requires a high level of automation, and we're now seeing the ability to bring AI in to help assist with that automation and operations challenge in running such a large open ran network.
Guy Daniels, TelecomTV (02:05):
Well, that brings us nicely to ai. What would you say is the role of AI in a telco network or perhaps what should be its role? Where can AI have the greatest impact and how would you quantify it?
Paul Miller, Wind River (02:20):
Yeah, so we're seeing it enter in a couple of different areas. We see certainly the SMO and the realtime Rick and these sort of things that are enabling dynamic control of the radio function within the 5G network to have interesting AI based applications. Things like dynamically controlling the direction of the signal based based on where the endpoints are and saving power and efficiency and that sort of thing is happening for our company. We're very involved in the infrastructure layers of the deployment of Open Ran and what we're seeing is open RAN is obviously founded on a cloud technology called O Cloud from the O Ran software alliance, and that architecture basically is Kubernetes deployed at an extremely high scale and that means you have both a disaggregated, what we call a vertically disaggregated system with hardware, a virtualization layer, and then the RAN application sitting on top of it as well as an east west disaggregation where you now have multiple vendors interoperating with each other through the open ran standards. That is a massively complex thing to deploy and maintain with multiple vendors involved in that architecture. And so we're finding applications for AI in the operations management and oversight of this type of network basically because the deployment of that system with the high node count, with the complexity of it, it's too much for a human being to absorb. You need software automation in order to maintain and operate that environment and AI is proven to be an incredible addition to that as we'll see today.
Guy Daniels, TelecomTV (03:53):
We'll come on to that in just a moment, but I'd just like to pick up and ask, given what you've told us about where you operate in the network and what you're seeing, how has Wind River been leveraging AI so far?
Paul Miller, Wind River (04:06):
Yeah, so we basically build a solution that we call studio operator that has a completely open source foundation based on the Open Infrastructure Foundation, Starling X, and that is a geo distributed Kubernetes solution that allows you to deploy a Kubernetes architecture across thousands of sites and run that from a single pane of glass. This is very important for an operator trying to run at high scale to deploy with zero touch provisioning, to maintain these virtualized applications across thousands of sites. On top of that, we layer on Wind River analytics that gives us the ability to monitor and visualize with the reporting and machine learning what's happening in that deployed architecture. Then on top of that, we deploy our orchestration and automation platform, which we call conductor, and this is where AI starts to enter into the application because conductor gives you a single pane of glass view over the entire network.
(05:00):
But as we had just spoken about, the ability to use AI on top of that is proven to be incredibly powerful. For example, predictive outage avoidance where all of the data is coming back in from these edge systems is aggregated in a data store. The AI has the ability to peruse that and identify problems before they actually occur. And so AI is bringing us the ability to identify a network outage before it occurs rather than having the service provider react to it after the event occurs. The other thing that it's useful for is debugging the network event correlation is an extremely important thing where you have all these different systems, there may be a fault that occurs that you see you experience an outage or an impact, but what was the root cause of that? Which system was it? Was it networking? Was it storage? Was it compute? Was it virtualization? And so the ability for an AI to consume large amounts of data and correlate that information and provide in seconds an output to the operator that says, here's exactly where your problem is that caused that outage is of massive value. So event correlation, root cause analysis, predictive outage avoidance and automation's operation are all areas that AI is being leveraged in the running of open ran networks today.
Guy Daniels, TelecomTV (06:14):
Incredibly useful. Now Paul, you promised us to show as an example, so perhaps you could take us through a demo now.
Paul Miller, Wind River (06:22):
Yeah, certainly. So this is a demonstration that we ran in Mobile World Congress this year, and this is the AI that we mentioned. It's a large language model AI sitting on top of the open ran network. It sits on top of that conductor element that I mentioned, and you can have a natural language interaction with the system. So here you can see the user querying how many sub clouds, what are the names of the sub clouds and the ability to gradually interact with it as you would talk to a human, are the systems running? Are there any faults in the system? Can you tell me about those systems? What are the name spaces that are deployed there into which I can deploy containerized applications? Are there any problems in the system? Here you can see what's the version of Kubernetes running. Perhaps you need to upgrade your Kubernetes and notice much like you would use modern AI tools like chat GPT, you're able to have a conversation with the AI that's now sitting on top of the network.
(07:21):
This replaces clicking on dashboards and firing scripts and driving API commands manually turning these hundreds of API calls into a simple normal English question that you ask the gen ai. You can see here the user asking about alarms in the cluster. Are any of the alarms critical and give me the details for that alarm. Notice how quickly you can tunnel down into the problems and resolve what they are. You'll see here the user asking, well, what do you suggest for me to fix this? How do I address this alarm? And the system will research that and come back in seconds with a variety of answers. In fact, we've demonstrated at Mobile World Congress this year, the AI can actually suggest commands that can be executed by the user to resolve the problem. We've seen AI able to analyze through log files and come back and point to the exact source code or script that's creating the problem that the user's experiencing.
(08:20):
And you can see here the power of having just a very easy interaction between a user and this massively deployed system that may have tens of thousands of sites asking questions and getting answers and helping solve problems. Right? Are there Kubernetes security certificates that are about to expire? Where are they? When's their expiration date? You don't have to know anything about the underlying commands or scripts or APIs to do this. You just have a conversation with the ai. So the things that you see it doing here as we're looking at would take a user looking up in manuals, researching through API commands or user interfaces hours to figure out how to resolve these things. And as you can see in this compelling demonstration, the AI is doing it in the blink of an eye as the user naturally interacts with them in normal manner. So pretty exciting demonstration and lots of videos like that that we gave examples of AI at Mobile World Congress this year. We think this is the next generation of operations for Open ran. The ability to operate a highly disaggregated distributed system through simple interaction with an AI is very, very powerful automation.
Guy Daniels, TelecomTV (09:31):
Fascinating. Paul, it's only when you see it do you realize the full potential. Thanks so much, Paul for that. A final question for you then. What should the industry expect from you and Wind River in the near future?
Paul Miller, Wind River (09:44):
Yeah, it's very simple productization, right? We're starting to see as Wind River is unique in the market of having the only highly scaled operational deployments of V Ran and open ran in the world. We're getting challenged by our customers in the market to solve the problems of running these systems at a high scale. So bringing the capabilities that you saw into the product and productizing them beyond proof of concept to actual product capabilities that our customers can consume and use is very, very important. And we expect to be bringing that to market this year. And then of course, expanding the role. We're actually bringing some of these AI tools into our own internal teams. The ability for our teams as they support the customer to leverage ai, to analyze logs, to look at source code, to rapidly turn around changes in addition to the operational features that you saw today. So it's both going to emerge in our products as well as being used within our teams internally to support our customers.
Guy Daniels, TelecomTV (10:41):
Well, thanks very much, Paul. It's good talking with you again and sharing your views on Open Ran and ai. Thanks very much.
Paul Miller, Wind River (10:47):
Thanks guy.
Hello, you are watching the Open Ran Summit part of our year-Round DSP Leaders Coverage. I'm Guy Daniels. Now as the development of Open Run continues and with commercial deployments now underway, telcos are investigating how to use AI to automate and optimize their radio access networks. Joining me now to explain more is Paul Miller, chief Technology Officer for Wind River. Hello, Paul. It's good to see you again. Let's start off by looking at the challenges that CSPs face. What are the biggest challenges as they roll out and start to manage their open RAN networks?
Paul Miller, Wind River (00:53):
Yeah, thanks guys. So it's really interesting what's going on now. We've reached a point in the evolution of open ran and virtual ran deployments where we now have several years of deployment at very, very high scale happening in multiple geographies and it's really exciting to see that because we're obviously seeing the fruition of many years of investment for many companies in that ecosystem bringing open ran to reality. And so the feasibility and performance are now proven to be superior than legacy ran, but there are new challenges emerging as we've gotten to this stage of maturity in deployment. Obviously it's one thing to build an infrastructure and an application that functions, it's quite another thing in a telco service provider to build this application so it can be maintained and operated at full scale in a telco service provider network. As we know, many 5G deployments have over a hundred thousand sites that are used for the VDU in the case of the open ran. And to manage that as a fully distributed architecture requires a high level of automation, and we're now seeing the ability to bring AI in to help assist with that automation and operations challenge in running such a large open ran network.
Guy Daniels, TelecomTV (02:05):
Well, that brings us nicely to ai. What would you say is the role of AI in a telco network or perhaps what should be its role? Where can AI have the greatest impact and how would you quantify it?
Paul Miller, Wind River (02:20):
Yeah, so we're seeing it enter in a couple of different areas. We see certainly the SMO and the realtime Rick and these sort of things that are enabling dynamic control of the radio function within the 5G network to have interesting AI based applications. Things like dynamically controlling the direction of the signal based based on where the endpoints are and saving power and efficiency and that sort of thing is happening for our company. We're very involved in the infrastructure layers of the deployment of Open Ran and what we're seeing is open RAN is obviously founded on a cloud technology called O Cloud from the O Ran software alliance, and that architecture basically is Kubernetes deployed at an extremely high scale and that means you have both a disaggregated, what we call a vertically disaggregated system with hardware, a virtualization layer, and then the RAN application sitting on top of it as well as an east west disaggregation where you now have multiple vendors interoperating with each other through the open ran standards. That is a massively complex thing to deploy and maintain with multiple vendors involved in that architecture. And so we're finding applications for AI in the operations management and oversight of this type of network basically because the deployment of that system with the high node count, with the complexity of it, it's too much for a human being to absorb. You need software automation in order to maintain and operate that environment and AI is proven to be an incredible addition to that as we'll see today.
Guy Daniels, TelecomTV (03:53):
We'll come on to that in just a moment, but I'd just like to pick up and ask, given what you've told us about where you operate in the network and what you're seeing, how has Wind River been leveraging AI so far?
Paul Miller, Wind River (04:06):
Yeah, so we basically build a solution that we call studio operator that has a completely open source foundation based on the Open Infrastructure Foundation, Starling X, and that is a geo distributed Kubernetes solution that allows you to deploy a Kubernetes architecture across thousands of sites and run that from a single pane of glass. This is very important for an operator trying to run at high scale to deploy with zero touch provisioning, to maintain these virtualized applications across thousands of sites. On top of that, we layer on Wind River analytics that gives us the ability to monitor and visualize with the reporting and machine learning what's happening in that deployed architecture. Then on top of that, we deploy our orchestration and automation platform, which we call conductor, and this is where AI starts to enter into the application because conductor gives you a single pane of glass view over the entire network.
(05:00):
But as we had just spoken about, the ability to use AI on top of that is proven to be incredibly powerful. For example, predictive outage avoidance where all of the data is coming back in from these edge systems is aggregated in a data store. The AI has the ability to peruse that and identify problems before they actually occur. And so AI is bringing us the ability to identify a network outage before it occurs rather than having the service provider react to it after the event occurs. The other thing that it's useful for is debugging the network event correlation is an extremely important thing where you have all these different systems, there may be a fault that occurs that you see you experience an outage or an impact, but what was the root cause of that? Which system was it? Was it networking? Was it storage? Was it compute? Was it virtualization? And so the ability for an AI to consume large amounts of data and correlate that information and provide in seconds an output to the operator that says, here's exactly where your problem is that caused that outage is of massive value. So event correlation, root cause analysis, predictive outage avoidance and automation's operation are all areas that AI is being leveraged in the running of open ran networks today.
Guy Daniels, TelecomTV (06:14):
Incredibly useful. Now Paul, you promised us to show as an example, so perhaps you could take us through a demo now.
Paul Miller, Wind River (06:22):
Yeah, certainly. So this is a demonstration that we ran in Mobile World Congress this year, and this is the AI that we mentioned. It's a large language model AI sitting on top of the open ran network. It sits on top of that conductor element that I mentioned, and you can have a natural language interaction with the system. So here you can see the user querying how many sub clouds, what are the names of the sub clouds and the ability to gradually interact with it as you would talk to a human, are the systems running? Are there any faults in the system? Can you tell me about those systems? What are the name spaces that are deployed there into which I can deploy containerized applications? Are there any problems in the system? Here you can see what's the version of Kubernetes running. Perhaps you need to upgrade your Kubernetes and notice much like you would use modern AI tools like chat GPT, you're able to have a conversation with the AI that's now sitting on top of the network.
(07:21):
This replaces clicking on dashboards and firing scripts and driving API commands manually turning these hundreds of API calls into a simple normal English question that you ask the gen ai. You can see here the user asking about alarms in the cluster. Are any of the alarms critical and give me the details for that alarm. Notice how quickly you can tunnel down into the problems and resolve what they are. You'll see here the user asking, well, what do you suggest for me to fix this? How do I address this alarm? And the system will research that and come back in seconds with a variety of answers. In fact, we've demonstrated at Mobile World Congress this year, the AI can actually suggest commands that can be executed by the user to resolve the problem. We've seen AI able to analyze through log files and come back and point to the exact source code or script that's creating the problem that the user's experiencing.
(08:20):
And you can see here the power of having just a very easy interaction between a user and this massively deployed system that may have tens of thousands of sites asking questions and getting answers and helping solve problems. Right? Are there Kubernetes security certificates that are about to expire? Where are they? When's their expiration date? You don't have to know anything about the underlying commands or scripts or APIs to do this. You just have a conversation with the ai. So the things that you see it doing here as we're looking at would take a user looking up in manuals, researching through API commands or user interfaces hours to figure out how to resolve these things. And as you can see in this compelling demonstration, the AI is doing it in the blink of an eye as the user naturally interacts with them in normal manner. So pretty exciting demonstration and lots of videos like that that we gave examples of AI at Mobile World Congress this year. We think this is the next generation of operations for Open ran. The ability to operate a highly disaggregated distributed system through simple interaction with an AI is very, very powerful automation.
Guy Daniels, TelecomTV (09:31):
Fascinating. Paul, it's only when you see it do you realize the full potential. Thanks so much, Paul for that. A final question for you then. What should the industry expect from you and Wind River in the near future?
Paul Miller, Wind River (09:44):
Yeah, it's very simple productization, right? We're starting to see as Wind River is unique in the market of having the only highly scaled operational deployments of V Ran and open ran in the world. We're getting challenged by our customers in the market to solve the problems of running these systems at a high scale. So bringing the capabilities that you saw into the product and productizing them beyond proof of concept to actual product capabilities that our customers can consume and use is very, very important. And we expect to be bringing that to market this year. And then of course, expanding the role. We're actually bringing some of these AI tools into our own internal teams. The ability for our teams as they support the customer to leverage ai, to analyze logs, to look at source code, to rapidly turn around changes in addition to the operational features that you saw today. So it's both going to emerge in our products as well as being used within our teams internally to support our customers.
Guy Daniels, TelecomTV (10:41):
Well, thanks very much, Paul. It's good talking with you again and sharing your views on Open Ran and ai. Thanks very much.
Paul Miller, Wind River (10:47):
Thanks guy.
Please note that video transcripts are provided for reference only – content may vary from the published video or contain inaccuracies.
Paul Miller, Chief Technology Officer, Wind River
As the development of Open RAN continues, and with commercial deployments now underway, telcos are investigating how to use AI to automate and optimise their radio access networks. Paul Miller, CTO of Wind River, discusses the role of AI in telco networks and where it will have the greatest impact. He also showcases a demonstration of AI in Wind River Studio Operator, and explains how it can be of benefit to telcos.
Recorded May 2024
