 Good afternoon. My name is Dennis Murphy and I work for IBM and software group and today I'm going to talk to you about automating the journey to cloud native intelligent networks So as we look at The modern telecom network platform the cloud has become integral to to every telco business transformation strategy You know as the world of network modernization evolves the growing focus is on open Hybrid cloud, you know compliant architectures So that services can be deployed as needed at the right hybrid location be it public or private cloud on the right hand side And at the right network location across across this picture if you like, you know And the move to network cloudification means that Network functions can run in the same cloud as IT and other workloads and this also means that workloads can be Automatically provisioned to run wherever they are best suited either near the the end user or you know, so-called edge computing are in the core data centers if you like and You know this evolution Offers the promise of greatly reduced cost of operation as well as you know a new level of flexibility and agility for for the telcos Allowing them to innovate and deploy new services quickly And as cheaply if you like as the competition and this is important because you know There's an increasing threat of increased competition, right? And this is lowering both revenue and operating margins at a time when all of us consumers Our data usage is going up, right particularly for for video trafficking video traffic, which is you know, it's exploding So the telcos and the CSPs need to quickly and efficiently scale the network capacity to meet these demands in a cost-effective way And you know we can no longer rely on traditional services to generate the revenue So this is why we need a modern network that will allow this communication service providers to to rapidly develop and deploy these new revenue generating services And that's that's all about modernizing the network by abstracting those legacy network appliances to a network cloud and This is all happening right now simultaneously with 5g the rollout of 5g deployments And to fully leverage the benefits of a modern network There must be a network and an automation strategy that is underpinned By artificial intelligence and this is what we'll talk about in the coming slides, you know complex our cloud networking And Services for large-scale the cloud native deployments are complex things to manage You know, there's there's cloud complexity, right? Services need to run across multiple age multiple public multiple private clouds that is tripled across that landscape The networking complexity, right? So the networking between these clouds and internet working within the cloud as well, of course But all the service components need to be designed and maintained across the network Dynamic services, you know that the services are much more dynamic than before You know location services Networks can be added can be deleted change deleted on on the fly in an autonomous fashion And this leads to you know operational complexity, right? This is you know, there's more effort up front today to design and test and it's critically important We do upfront testing and design of this In particular for day two operation use cases, you know once services are ploid and into production And you know, there's an increasing complexity. This is heavily increasing as well So we need to figure all of that out and you know traditional tools You know, they're more manual based and they require static programming And they're not really suited to this this new modern telco environment to this new environment that we're looking at Complex dependencies, you know, they exist across multiple technology layers, you know, if you look at the Right up and down the disaggregated stack where we now have open hardware and open controllers open nickards Etc right up to the network functions from multiple and different vendors Up to the stack to the design of the network services themselves That logical network design and stitching across those sites like I mentioned So these complex dependencies across all of these multiple technology layers You know analogy would be like the enterprise edge cloud IT and the networking workloads they must cooperate across these independent layers and Connected locations to deliver that end-to-end service, you know And each layer here is designed and managed independently of the others and delivering a service if you like to the layer above You know for example a an IT application owner might be designing a video surveillance application And you know, they need low-level layer to layer tree networking details Would be required to deliver that application traffic But that's abstracted from that IT application owner who will work with the network engineer You know who can to design the required network design and you know It uses orchestration design tooling to design out the networking that that IT service requires You know so these logical network services in our dynamic and like we said that can be added and moved and changed locations You know for example of that video surveillance application was sitting on a pod over here and some AI infused Monitoring suggested that you know move it to another pod over here as we know pods go up and down quite frequently That impact and the change of that IT application from here to there has an impact on all the networking service below that All of the layers up and down the stack and they need to be managed that ripple effect across the network needs to be managed accordingly and If we look at this, you know from a top-down point of view like we said on the network service point of view, you know What what what the questions we need to answer or what where do we put which version of this network function? What existing network functions exist that we need to bind to? When should a network function move from one site, you know along this this slice here to another site So that network service changes all of these changes Cause a ripple effect across all the layers across multiple different cloud locations You know and these changes need to be automated to cope with new levels of operational complexity You know an example here with this this slide would be to know a 5g network service You know can share and create dedicated network functions across many network cloud locations we can also look at this then from a you know a bottom's up point of view if you like from a hardware tuning point of view and You know we can look at this from the hardware up through the the the esteem controllers up through the leaf spine Up to the underlay overlay from that point of view, you know gone are the hard distinctions from the past right between infrastructure Between infrastructure network functions and the end-to-end services These are now replaced by a flexible hierarchy of interconnected software services up and down the stack You know virtual network functions are cloud native network functions VMs and containers You know the all these things now require specific hardware and tuning things like nick parameters things like hypervisor parameters Things like container networking plugins You know etc etc so all of these The design of the network service and the underlying you know the network functions are composing up that network service put requirements on those layers below like we said and Tuning of the specific hardware and tuning of the specific controllers, etc. Which is all important to take into consideration as we do our our network design so Ultimately all of this is about the move to putting software at the center of the network Know the way to really achieve both the innovation gains promised by cloud network cloudification and 5G and To maintain operational costs by is by adopting a cloud software culture And this starts with software management approaches and adapting them You know adopting these software management approaches and adapting them for the telco So let's expand this in a little bit. So on the left hand side here You know we talked about a standardized life cycle You know a standardized life cycle model and API allows Dynamic assemblies of network functions and applications. This lends itself to lower complexity You know everything abiding by this standard life life cycle model allows you to layer huge amounts of automation on top You know all network software function components Implement these standard life cycle models and API's then the engine itself can figure out The life cycle tasks in any complex service required to keep the inserts to keep the entire service In an active mode or an active state and doing this without any upfront programming You know, it's all about making everything look the same all the components must implement each standard life cycle transition Then everything looks the same So it's really about modeling dependencies and relationships and You know, it's not about workflows or upfront workflows that try to anticipate every possible state And then we have the concept in in cloud made up here of this intent engine So this declarative based approach so the intent engine can use these relationships Then these dependencies and relationships to figure out how to get into the desired states This is my design is where I need to be and we can automatically figure out how to get there You know, it calculates this intent this engine its intent base And what we mean by that is that it calculates and executes the minimum set of required actions Considering the actual services topology at this moment in time to reach the desired state or the design to be state for the target service instance And this in this way we can hide all the underlying service complexity So that you the service designer can focus on designing your service and what that service should deliver to your end customer to us as consumers And not how it should be deployed or implemented at the detailed infrastructure level Like we mentioned earlier around those the different layers of the stack You know these network functions put dependencies on different layers of the infrastructure and Then the engine itself, you know, we've got a set of opinion patterns that can support the healing of broken services and resources Etc. And all of this is done, you know You know things like, you know, it's done in a process execution plans that can resolve Placement strategies can resolve any shared resources things like that The analogy that can be used here similar to a sat nav system where a user programs their destination and the sat nav figures out The best route to get there depending on if there's a crash and it can reroute you automatically, you know If you're going from corp to Dublin and there's roadblocks or whatever We figured out the next possible best possible route to get you to your destination You know in the same way this intent based solution can be loaded with a set of simple service models and All the operational processes then to require to keep that service optimized are in place And then when we look at the cloud-based tool chain, you know to scale any cloud-based networking program You know a unified operations and network engineering model is combined with a set of automation tools that can simplify and Automate the complexities of the end-to-end lifecycle of the network functions themselves and also the network service And You know you need to look for a solution that's got in-built service behavior testing tools No solutions that allow you to be to fully test the network function the onboarding cycle lifecycle Fully test the network function to see that it does it when it says in the tin and do all this in an automated way Now traditionally these tasks are very time-consuming with programming having to reprogram and run all the tests again for say a New version of a network function came from from the third party But you know in a cloud native fashion in a cloud operating model You know an automated test framework allows you to spin up entire dev environments This is taking advantage of the cloud You know install and activate all the network functions and network services on those environments Add additional test resources things like traffic generators or metric generators or you know and run the service through its entire lifecycle And these tests and can be part of an automatic, you know Continuous integration continues to be a really pipeline, you know part of ready-for-service tests You'd be like really old the whole process of upgrading to a new version of a network service or patching in a new version of a network function You know kicks off that automatic test. It wants through the entire lifecycle and Check the box and then you know performance is all good security is all good You can then automatically, you know de-install the service from the test environment and push it to the production environment So that's what we mean by applying a lot of these cloud native techniques that we have learned from the IT world And applying machine-enabled automation here with the intent engine that you know, it's all about these opinionated Patterns that model dependencies and relationships and not trying to figure out everything up front Let the machine take over and figure out the best possible pattern So once then we have All of that set up, you know, we have now moved our workloads into production It's very very important then, you know for you know How do we manage or how do we operationalize the data to operational use cases? No AI and it's a term used but AI for network operations is all about the infusion of AI to provide operational efficiencies such as predictive alerts, outage avoidance, instance reduction, etc, etc And into the network operations And the journey to AI ops has a number of different maturity levels And if you look down here on the bottom left of the There's kind of the middle of the chart here, you know, it's simplified maturity levels So you know noise reduction, de-duplication of events, de-duplication of alarms Then we move on to more reactive, you know real-time insights into what the data trends are Correlations things like that then we move on to more of a predictive real-time dynamic insights for Probable cause identification things like predictive correlations. This is what the the normal behavior is Multivariate colorations all sorts of machine learning can be applied here and then going to more proactive Maturity level, you'd be like so from smart incidents avoid outages get ahead of any problems before they come customer affecting And all of these maturity levels, they all follow the kind of the similar four stages of Of the AI ops journey, you'd be like so it's collecting your data that provide the relevant Data insights from KPIs from events from time series data Organize that data so curated cleanse it Govern that data if you like Analyze the insights, you know apply machine learning apply AI to understand what are the Insights we can gather from the data and then infuse that AI into operational processes into the closed loop, etc and then as we Talk about implementing AI ops, you know, it's important to Look at this from a perspective of you know to implement intelligent operations, you know You need to tie all the dots across all of the data So you need to tie signals across structured data and you need to tie signals into and unstructured data for multiple sources like we have on the left hand side Up and down our stack here But we tie that data from the structured and the unstructured data together So that you know we correlate all of this data together Right, so we tie signals from from multiple sources and we can provide a clear view Of anomalies of clear view of linkages to sources for faster investigation faster resolution Um, and there's there's we can derive hidden insights from all of this data that we're collecting No structured data, you know things like events time series data KPIs performance management data but equally then unstructured data things like logs tickets Manuals user manuals for different types of network functions for example coming from the third party vendors You know in order to leverage AI and sophisticated analytics You know, we need to have a solid foundation like we have here with robust tooling Probably all of this type of data this was structured and unstructured data and we need to understand the data You know all of this data together is really really valuable for us in network operations You know, but we need a real time. What's a ground truth, right? What is the state of the what is the state of my network for managing? You know that ground truth can help us manage these complex applications these complex services these dynamic services that we're rolling out um And you know these tools allow us to you know anticipate and address risks proactively like that maturity carbon AI, right? You know draw insights from more complex unstructured data Get ahead of any problems and all of this lends itself into you know towards automation How can we automate this you know building that foundation? Enrich a bit topology data enrich a bit AI other sophisticated toolings And then that can further reduce cost further reduce risk You know and I will you know allow us to have more sophisticated strategies, which we can discuss Next here so designing networks for the cloud with AI and automation so Expanding all we just previously said You know a cloud native approach that encourages reuse is really important Now this demands intent-based definitions that let the machine if you like handle the specifics You know a lot of this is about build to manage You know these automation models right up and down the stack we talked with them earlier And so these intent-based models You know make these available not just for orchestrating and configuring and instantiating day one day zero and day one Of your network service, but make them available to assurance as well So we can derive additional insights out of there So this set of predefined lifecycle types these dependencies and relationships that we spoke to And you know have a set of automation artifacts If you like a library of opinionated patterns That can address planned and more importantly or equally as important unplanned And you know changes in the network so for service service restoration use cases for example You know an assurance in the eye is monitoring everything On the network at this stage now as we see here Let's get this through a little bit You know we're looking to you know An infrastructure event can cause an error for example Uh cascading of errors no assurance in the eye help us to Select the appropriated opinion of patterns that addresses for unplanned changes There's a series of feedback loops back into the orchestration layer. Um, all the lights could be green for example Um users could be complaining on social media, but everything in the north looks looks good, right? Um, you know, we've got a client that has um Twitter sentiment analysis overlay than the knock and that helps the prioritization You know, this is in in countries who have very high social media And you know users are on all the time, right? So other customers have weather weather analysis feeding into the knock So all of these insights can help with the prioritization You know of reconciling the actual state Of the and the target state of the actual network cloud stacks across these different locations Um, you know, so sense and respond to issues to opportunities for optimization And it's worth reiterating here as well that you know This is all done with these models based approach without any upfront programming So we're you know, we're making the same automation models available for instantiation Right up and down the stack from the nfei up through the network function up through the network service layer For the enterprise application there available for assurance and for ai monitoring And Along with that then, you know, we require To leverage a cloud operating model. Um, and what you know We should understand that the the need to those techniques I spoke to a while ago that they need to adopt this software development management. Um Techniques we also need to you know, reorganize ourselves as well We can't simply try to all automate existing legacy processes You know these processes were designed for Networks that were based on hardware or change in configuration and you know, it's slow, right? And a lot of it's often manual, right network upgrades are measured in in months, right things like that You know existing our legacy processes and tooling. They've also got a high degree of manual touch You know that limits the ability to deliver on demand or self-correcting cloud services You know most operating centers, you know require skilled technicians who could perform mainly manual tasks So we now have this cloud networking automation platform And that we can embrace, you know and learn from the IT in the dev ops world dev ops movement if you like And enable complex network services to be designed, created and continuously optimized across Hybrid and across public private and across these distributed cloud locations And this requires new tooling process and skills and culture, right? You know, if you think about what the Um, and then the life of an ops person is in this type of cloud operating models You know, it's things they configure once and then you can apply that multiple times in a cookie cutter fashion You know all changes in configuration management are through these two character templates You know engineers don't need to tweak low level system configs directly You know drive your top level intent and that will then try the low level commands and you know conflict changes Which lends itself into proactive Close to control that classic observe heal test You know and this network dev ops environment It really enables a lean and effective way to manage and to and to roll out faster implementation of function functionality And you know, but to manage the complexity of that environment and ultimately to improve customer experience In summary then or to wrap up, you know, we can talk The move to this cloud native converged IT network cloud, you know It can really be summed up into no apply intelligent automation And we're talking about this intent based modeling And focus which is focusing on exposing and delivering what a consumer wants or what a customer wants not exposing the technology Kind of how do I how do I deliver this and then applying this to the operational changes? You know, adopt those new processes adopt those new operating model Infuse with AI everywhere if you like You know, we in IBM are helping clients to safely adopt this revolutionary approach To a software based network management, you know We're allowing clients to to benefit from the what we've experienced for many years now from the already demonstrated benefits of cloud native You know learning from other industries where there is high innovation rate where there is lower You know, we've proven that we can lower the operational costs You know as an analyst said to me earlier in the week Now ultimately 5g is about the move to putting software at the center of the network And the approach then our approach in IBM We believe enables that open open ecosystem for delivering that open hybrid network cloud You know, if you want to find out more you can take a look at the IBM virtual boot At the conference I can reach out to me directly So thank you for your time today. I hope you found this informative and available for any questions. So thank you