SR/SRv6 Crosswork innovation​

Hello everyone. Good morning to those joining from the Americas and good afternoon to those joining from the European region. Welcome back to the.00:103rd session in the series of path to autonomous network series. The today's topic is on SR SRV six segment automation with crosswork network controller, like always, this is a this we can we'll try we'll keep this session as interactive.00:20As possible, so if you have any questions, please feel free to raise your hand and I will unmute you to to ask your question live. Today's today's presenter is Eric coming from the TME group, and also the demo specialist is.00:40J also coming as the TME are experts from the TME group, and the panelists on the, on the session are the customer success specialists who will be helping you answer questions any, answer any questions you may have. The chat window is also open for you to post your question there.01:00And with that if there's any, anything, any questions or any doubts, like I said, please feel free to raise your hand or post your question on the chat window and our panelists will help you answer those questions. With that, I'll hand it over to Eric, enjoy the session. Thank you.01:21Thank you. Welcome everyone to this session. So as explained, we, we would like to spend the next 90 min to discuss about SRMPLS and SRV six automation. Why is it critical and what are the tools that.01:41CNC, the crosswalk network controller can offer.02:01To simplify this SRS SRV six automation. You will see, so we will speak a little bit about SRMPLS, but we we will really focus on SRV six, not only because we think it's it's really the.02:05A real new solution which is good for many many use cases, but also because it's coming with new, new tools to to simplify automation and operation. So we will spend some time to to discuss about those and to give you to give you demos.02:21Of of this innovation. So I'm here today with Sujay Sujay is our technical marketing engineer.02:41And will help me on the on the demo side and he will cover some, some of the topics and I'm Eric or two, I'm one of the crosswork network controller product manager. So what we would like to discuss today we will start.02:49I know it's it is a 3rd presentation in the same same group so you should be aware now of what is crosswork, what is CNC, but I will do a very quick refresh. And then we will jump into the, the meat of this of this session on SRMPLS and SRVC.03:05What do we do today? Where is the innovation and where are we going for around those topics? And one of the key topic for when we speak about SRV six in particular is really what we the bandwidth management. What can we do to manage bandwidth SRV six bandwidth in a efficient way?03:26And you will see we will discuss the strategy and you will see that we have three pillars of that strategy that we will cover, so observability, planning, and congestion mitigation.03:45Before we before we wrap up, we will also discuss about SRV six migration and how crosswalk and the crosswalk tools can help you or your customer to migrate, to migrate to SRV six. It may it may look scary like this, but.04:00You will see that it's not that it's it's not that complex and that we have tools to simplify the migration, the migration process. As said, we can keep this session interactive so don't hesitate to ask questions either in the in the chat window or in in.04:18Live if if needed. And yes, I think the session should be recorded.04:37But I will let Babana confirm. The session is recorded and yeah. I saw the question on the chat so that's why I was, I was replying. Okay, so just a quick refresh on on crosswalk and so crosswalk is a is a is a.04:44Value of products and crosswork network controller is one of them. And the idea is really to to.05:03To pro to simplify the life of the, of the network administrator and to simplify the automation of multi domain multi vendor networks. And we do it around three pillars or three activity action.05:10Visibility and insights, and this is a, we see this is a closed loop between those three activities if you want and we are adding more and more AI agent capabilities to those to the existing features to to make it even more efficient.05:30And if you attended the previous presentation on autonomous networking, the goal is to go more and more towards autonomous networks, networking. When we speak about CNC and automation, we are talking about service orchestration. So everything around device configuration.05:47And service provisioning.06:08We are talking about automated assurance, once we provision a service, how can we assure of this service and verify that the intent is is met? Optimization, this is really critical when we speak about SRMPLS SRV six, make sure that we are.06:10Listening to the network events and react in real time to reoptimize to reoptimize the network if needed. Visualization of course and we'll we'll discuss more in detail about it but.06:30If you are familiar with CNC, you know that we are spending a lot of time to show services, to show topology, to provide as as as much information as we can in the, in the UI so that you have a clear view on on your network. And you will see that.06:44In specifically with SRV six, we have new tools to even provide a better visualization of the traffic and, and the traffic flows which are pretty useful. Workflow so it's all about workflows, so you once again if you are familiar with.07:02You know we have a bunch of workflows embedded workflows and we have also a workflow engine that that can work in coordination with with CNC.07:22And the last activity which should be the 1st one, by the way is usually the planning and we have two, two solutions, either the crosswork planning for offline planning, which has been around for, for, for many years and which is very powerful to to simulate the network.07:31And simulate what if what if scenarios, and you will see that with SRV six, we start to introduce some.07:51Level of real time capacity planning to that will be less sophisticated than swalk planning but that will give you a quick sense of where could be the issues in the network, what could be the hotspots or or the bottlenecks.07:59So we are going to discuss all of this. Quick recap on CNC. Once again, I hope most of you should be aware of what crosswork network controller does and it has been around for many years. We are right now in version seven to two and about to release.08:18Version eight.zero of CNC and what you will find in CNC is obviously device life cycle management, service life cycle management including provisioning and visual visualization and assurance, real time network path computation and optimization.08:38We do network slice automation we start to do more and more closed loop process automation, and as I mentioned earlier, we are adding multi agency we we are we have this multi agency framework and we are adding.08:58Multi, AI agent use cases on top of CNC as we move forward in a less than a minute, so very, I hope it's everyone is familiar with the with the product now and let me now jump into the.09:13The the SR and SRV six discussion more in detail.09:32So, why, why do we why do we spend why do we want to spend some time on SR to discuss about SRMPLS SRV six automation? Basically I mean, if you think about it SRMPLS and SRV six provides already a.09:37I would say a programmable data plane, but now what does it mean to to automate and why do we need to automate it? And typically, you want, you want to automate it for many reasons. The 1st one is speed. You want to be able to provision services in a very.09:54Quick way, and you want to remove as many manual touchpoints as you can. You want to do it at scale. Usually I mean, depending on the size of the network or of your network, you may have to provision many multiple services at the same time and you want to do it.10:14So you want to do it fast and at scale. You want to do it in a reliable or reliable manner, so make sure that you can validate what you plan to provision and see what it looks like once you provision your your services. And of course you want to keep some agility.10:34Or some, let's say reactiveness, so once something goes wrong in the network or if the traffic changes or if anything happens in the network, you want to make sure that the controller is going to react and to to adapt if needed to to deliver what you what you were expecting at the beginning.10:54So, the idea is really that we start from the intent, so you describe which kind of service you want, which kind of parameters, which kind of constraints, and you want that once you provision this service or those services at scale, the system will will manage them properly.11:13Over time, which is basically a kind of a loop where you start from, from the intent, so you define your service objectives, your service constraints, you, you're going to to that the system is going to translate that into a a.11:33Policy, so keep in mind that the controller knows the topology, the current network topology and especially when it comes to SRMPLS SRV six, the existing policies and and existing tunnels and.11:52The service is going to be provisioned with a potentially with a new tunnel or a new, a new policy that that respects the constraints and and the objectives. And once we deploy this, you want to make sure that I mean, 1st of all, the system is going to take care of the of the provisioning of the different nodes.12:10The system is going to take care of the assurance.12:30Of this service those services and those policies and last but not least, the system is going to start op optimizing all the time, I mean as the network leaves, make sure that optimization is running all the time to to keep the intent.12:32As it was designed initially.12:52So that's that's really the the life cycle of of the SRSV six automation in in in CNC. And you will find this for both for SRMPLS and SRV six. If I focus little bit on SRMPLS itself and once again this is not where we are going to spend much more time.12:55But just a recap maybe of what we do with SRMPS today in CNC starting with SRMPS topology, so in CNC you will find of course the network inventory, the SRT policies, flex algos, the service the services that are, that are provisioned over that SRM.13:15Infrastructure.13:37So everything is is discovered or or part of the CNC database and you can, you can visualize those on the map or expose through the APIs as, as you, as you wish. Service provisioning is a key activity for, for CNC as you know. So making sure that you can provision layer two, layer three services.13:39In a very easy way, in a multi vendor environment and with SRMPLS steering and as you know probably we are leveraged leveraging this NSO provisioning engine, which is embedded into into CNC.13:59And which gives you a lot of flexibility to provision different types of services with with different vendors. That's one of the core activities of CNC as well, making sure that we can compute path with the explicit constraints.14:15You know constraints or objectives, by the way, flexalgo.14:35Path diversity and and so on and so forth. So that's the the traffic engineering engine and of course you can stitch your services to to the right to the right policy. Once policies are deployed, one of the thing that.14:39We can, we can do, is to manage bandwidth and one of the 1st activities that we have is really to manage congestion the risk of congestion let's say through mitigation and we'll discuss a little bit about this in detail.14:57What the what do we mean by local congestion mitigation today? And some of the most recent innovation that we we introduced for our own LCM local congestion mitigation in the late, in the last last CNC release. We have also some bandwidth management.15:15Active feature per se, like bandwidth on demand or circuit style circuit style being a kind of a guaranteed bandwidth services. And last last in my list but not least of course service health, which is all the service centric assurance that we we provide so when you provision.15:34The service should define how this service should be monitored and what what the service health means and the system will monitor the will monitor the service in real time.15:54So just let me spend a few minutes on on SRPS congestion mitigation techniques and innovation and what we call LCM and by the way, this is.16:08Available today for SRMPLS it will that will be available in 2027 for SRV six as well. So we'll come back on this specific aspect of SRV six when it comes to consistent mitigation, but keep in mind that.16:24The behavior I'm going to describe is, is more or less will be available for SRV six as well. So the idea is to detect I mean the system is going to detect congestion or the risk of congestion let's say, so we monitor some interfaces.16:41And, when we, when the the optimization engine is going to detect that there maybe a risk of congestion, it's going to compute some alternative path.16:59That would, that would be, that could be used to reroute some of the traffic and to remove the congestion. There are two, two aspects today, so this computation is done automatically.17:15And the outcome is two things. Either we do it, you can select the manual mode so someone is going to say, ok, go ahead and provision those what we call tactical policies to remove the postion. And now since seven CNC seven.two.17:30We support also an automated mode so you can leave the system to to manage the congestion automatically. And we'll I'll come back on this new mode, I mean this new automated mode in the next slide.17:48In which in both case those tactical policies will be deployed, so the idea is to is to deploy those tactical policies and and so that the the traffic is is being rerouted and the excessive excessive traffic is being rerouted and the congestion is avoided.18:04And depending on how it looks I mean how it it the the traffic behaves after after the re rerouting, the system will propose further further actions, maybe to come back to the normal or or or or or things like that.18:26What so as I mentioned, what we are, what, what we did in CNC seven.two is we need to add this automated look LCM mode. So in the past, you need someone to approve the tactical policy creation in this mode, in this automated mode.18:46This can be.19:05I mean the system can decide and and you as a user you decide if you want to do to use manual or automated of course not the system is not deciding for you, but if you select that, if you decide to use the automated mode, then the system will will will will do it for you.19:07And we provision automatically the the tactical SR policies. What we did, we did two things to to go with this feature. One thing is to support what we call multi segment list and I will I will go into the detail in the next slide and also to.19:26We we enhanced the history, the LCM history, so we used to have some history in the past, so you could see what happened, but in that case we since there is a lot of automation, we decided to to make it more complete the history more complete so that you can really track what.19:46What was the the congestion situation and.20:06Which decision were made at that time and what which actions was were taken or taken, sorry. And and once again the goal, the goal of all this is really to avoid congestion as much as we can, and so that make sure that there are there is a minim minimized impact on, on service, of course.20:11And optimal optimal utilization of the network.20:30So I mentioned earlier that we we are to to support this automated mode or to go along with this automated mode, we introduce the support for weighted multi signal list, and basically it allows us to.20:34Have multiple paths in one under one SRT policy, and it was a need for to create those tactical policies, and in the past we were using what we call parallel parallel T policies. We were using multiple policies to build.20:51Those tactical but that was not very efficient and that was not very easy to manage. Now, the good thing with multi segment list is that we have only one single SRT policy with multiple paths and the traffic will be distributed around around those paths.21:11It's what it provides is 1st all it provides a more granular traffic distribution but more important probably it's it simplifies.21:32The the creation of the SRT policy and how we manage it and it avoids any issues with, with, with the ECMP ECMP effect. So multi segments can be used in different.21:44Use case but the one of the main use cases for us those those tactical policies creation. Just a quick note at the bottom in still roadmap at this stage, but we are very close to release CNC eight.zero, and in CNC eight.zero we will we will.22:01We provide what we call selective deployment of tactical policies.22:21Sorry, which means of course it's only applicable to manual to manual mode, but in because and in manual mode you will be the the user will be will have the choice to select which tactical policy he wants to create. So the system is going to compute a bunch of tactic of potential tactic.22:26Called policies and the user will be able to select which one he wants to to to enable or or on the other hand which one he doesn't want to enable.22:45And the system will will always recompute to make sure that that's after the selection the the solution still still works to avoid congestion. So that gives you a lot of flexibility to select which, which policy to create, which tactical policy to create and where when.22:55Which one should be should be instantiated.23:17Any question around SRMPLS automation? I want I was I went pretty quickly but would like to spend some time on SRV six? Yes. If there's any questions now is the time, please raise your hand and I'll unmute you to ask your question live with Derek.23:22Okay, we have a question on the chat window Eric. Is the. So I for some reason I see only the 1st part of the question, but if I if I understood correctly the question.23:41I think the question was, do we manage the the provisioning of tactical policies fully in a fully automated way? I'm sure if it was a question but right now you have two choices. Either you say you do it fully manual manually or fully automated.24:00Yeah, I don't know if that answers the question let me try to see if I can get the full question in a simple way.24:20Yeah, I think you answered the question. Yeah, so we have I could not see the complete question. So yes, if you go automated, you everything will be taken care taken care by by the system.24:28Okay, let me anyway, you can, you can ask question anytime during the session. Let me just move forward and start discussing about SRV six automation. There is a lot of interesting stuff going along with with SRV six automation. Just once again.24:46Quick summary of the use cases that we support today or s for some of them that will support tomorrow. Once again SRV six topology we are talking about the network inventory, all the SRV six infrastructure, so the T policies, the flex algo, the services.25:06Everything in, is managed by CNC as usual. Service provisioning, just like SRMPLS you can you can provision your layer two, your layer three services with SRV six steering or flexalgo steering so it's up to you. Traffic engineering.25:26Same we have past computing past computation with explicit constraints or or objectives orgo. What we are going to introduce in with SRV six.25:46Is observability. So we'll come back on this in detail, but typically you will see that we are introducing new tools in particular what we call demand metrics and routing analytics where we can have a.26:02A very detailed view of of your path on one side and very detailed view of the traffic demands, so the the traffic flows in the network. And that's something that was not easy to do in the past with with our SRMPLS.26:18Now it's become much much much easier with SRV six, so we are going to discuss this in in detail. And the good thing with with this observability feature that it opens new, new use case as well. So that's what we are going to discuss. One of the use case being postion mitigation, so once again we we.26:37You will see LCM.26:57As just like for SRMPLS, but we are also introducing new what we call global global or centralized mitigation solution that that are quite interesting. Service has as just like SRMPLS we can do a service else for.26:59SRV six based services and migration I mentioned at the very beginning and where Sujay will present what we can do around the migration to to SRV six. So I believe it's quite a comprehensive list of of features around that and capabilities.27:20Is around six automation and you will see that we are keeping improving those those features. So if we try to summarize where we are today and where we are going, on the left you have a list of of.27:39What we called let's say foundational features for SRV six automation, including network discovery, including the IPv four IPv six topology visualization, SRV six infrastructure visualization, the T policies, the flex algo, the services.27:57And so on and so forth.28:17Moving forward and we started in CNC seven.two earlier this year, we are introducing a 1st level of SRV six observability and once again, my apologize if you've never heard about deterministic demand metrics before, but we are going to explain, I'm going to explain what, what we mean.28:20By that, and routing analytics, so those two new features were going to spend some time and they were introduced in the CNC seven.two. And moving forward.28:39And I will share a more more complete roadmap at the end. We are con we are going to continue to enhance the observability to add some planning capabilities and to work on the mitigation aspects. So.28:49Right, right now we are in the green green area and we have a lot that we plan to to do for for SRV six automation and operation. Just a few screenshots to to give you a sense of what you get today if you use CNC with SRV six.29:06Once again it's things you should be familiar with but you can see the SRV six policies ECMP path, and when you have if you pick one SRV six policy, you can drill down to the policy detail, the headlines, the endpoints the color.29:26As binding seed, and you can also get a view on the candidate pass details. So a complete inventory of your on which you're going to build, you build your services.29:45Okay see there was some questions so I can try to.30:10So 1st 1st question was about the last question was about the date for CNC eight zero or right now we are targeting December 26 for CNC eight zero. And there was another question let me read out for you Eric.30:16Which of these use cases you described with SRV six feature support are not possible to implement with SRMPLS but with SR features? Yeah, let me just go back to the slide. You will see, I mean, you will see the during the presentation anyway, but what what we what we can do in a.30:34In SRV six is that we leverage, we leverage what we call the SRV six locator counters to build the demand metrics. So basically we will be able to, to monitor the traffic from between two nodes, between two locators if you want and to build this what.30:54Because it's demand metrics. If you want to do it with SR and PLS or any other networking technology, it's more complex. I mean we do it. Typically we do it like in crosswalk planning, e.g., but based on netflow.31:14Thing like that and as you know, probably netro is pretty heavy to, to, to put in place and you need to sample it where with the locator counters there is NO more sampling. It's hardware, hardware based. This is streaming telemetry so very efficient, so.31:28That's really something we can do in SRV six that we don't, we are not able to do easily with SRMPLS. And once we have this accurate demand metrics, you will see that it opens new, new use case, so what we what we call automated capacity planning ACP will be based on.31:48On this DDM on this demand metrics, but what we call the the global the global position mitigation once again will be based on this demand metrics. So.32:08Being able to do to to keep to or to build a real time accurate demand metrics in the network is, is enabling is enabling new use cases like capacity planning consistent mitigation and probably much more.32:21And of course, I mean and then there are some some of course some differences between SRV six and SRMPLS. I mean some benefits regardless of automation but I'm I'm not going to to discuss about those but clearly SRV six is is really powerful and.32:41That's why we expect to see more and more ramp up around SRV six. But automation and those new innovations that we're going to discuss in the next slides are really one of the aspects that that make that makes SRV six interesting.33:00So I hope it answers the question but we can we can let me finish let me move forward in the presentation and we can maybe we can re discuss at the end we will see the demo as well so ok let me go back to the slide very quickly yeah one more question.33:19When is the CNC eight.zero planned? The release?33:39Okay, I thought I replied to so we we plan it for 26 December. Okay so let's, we're going to spend some time on on what we call the SRV six bandwidth bandwidth optimization or bandwidth management.33:43Strategy and the three pillars that we see for this for this strategy. And those three pillars are observability, planning, congestion mitigation. When we say observability, I'm.34:04So of course it's it's we mean all the all the information that we got excuse me, the network topologies, the the policies and so on. But more important is really what I was trying to describe with the demand metrics which provides a real time visibility into the live in the.34:18Into the actual traffic flows, and when we combine this demand metrics with some what if scenarios.34:38We can we can we can we can identify the congestion hotspots or or bottlenecks. So the idea is really to combine the topology, the demand metrics and the events that we receive with CNC to provide the complete visibility.34:49On the, on the SRV six network and the behavior.35:09That will help you or that will help your customer to optimize the bandwidth utilization and of course that will help you to identify any risk in the topology or in the routing that you should.35:14You should pay attention to before before enabling services. So that's where we need some planning and as I mentioned, we can do offline, very sophisticated, very advanced planning with crosswork planning or you will see in the presentation that we can do.35:32More real time more real time kind of of capacity planning which is still interesting to identify potential hotspots and potential gaps in the network. And the goal of this, of this planning exercise is really to avoid as much as we can the congestion situation.35:52So we don't want to go.36:11Into a situation where where we have congestion because it's not easy to fix. It's not easy to mitigate. But sometimes even if you plan if you do a very good planning, you may, you may have issues or you may have, I don't know things bad things happen in a network you can have link failure.36:14Or not failure.36:33You can have access traffic, whatever. So, even if you do your planning properly, you may have some congestion happening and in that case, your expectation probably is that the system will react and will will will fix or will how to say, will mitigate the.36:35Either automatically or or manually. And the goal is to to once again, even with SRV six will be to use tactical T for congestion mitigation. There, so we will have the same mechanism that we discussed for SRMPLS with local congestion mitigation, which is local.36:56By definition, but with SRV six, thanks to the demand matrix, we will be able to be smarter, let's say.37:16When we reroute traffic, we will be we will be able to reroute traffic to to the let's say the shortest or the most efficient path to the destination. Since we know where the traffic comes from and where the traffic is going, it will be easier to to reroute in a, in a good.37:25In a very efficient way. So let's start with observability and discuss about two new tools. 1st one is the routing analytics and 2nd one is DDM that probably mentioned ten times now. So.37:45Let's start with routing analytics, and the idea of routing analytics is really to show information about the path, about the ECMP path and and and the the performance of those ECMP path. So the idea is to provide path visibility so you can pick two.38:01Nodes, a flex algo and the system will show you all the paths, all the ECMP paths between those two nodes.38:21And routing is also going to track changes, so how many times the path has changed? How many times the paths were stable so that you know exactly how the how the network is behaving? Is it stable? Is it changing.38:30Switching paths all the time it can be interesting. And then when we combine those path information which are, let's say topology information.38:50With performance information like live ne liveness loss or latency, you have a pretty complete view about about your network and about the path. So that's what we did in CNC and the idea was really to to provide this real time path information.39:00Including all the CMP paths between two given nodes, and as you can imagine, it will open some new use case like service troubleshooting traffic engineering when you when you need to design a new, a new policy and so on and so forth.39:21So Sujay will present a demo where we have a 1st use case in CNC of where we use routing analytics to to to show ECMP path for a given flex algo.39:40And just I did not explain initially but so just if you are very familiar with CNC and you know how to look at the CNC components, you will see that routing analytics is a new or path analytics is a new is a new component under the, the, the component family. Just to explain how it works.39:56Very quickly, in fact it it listens to the BGPLS feed from the network. So that's where it, it knows everything about policies path in the network. And it will compute the path once you give a pair of nodes and a flexalgo.40:16And of course all this information is available through the API and just if you are familiar with what we call crosswalk assurance, Crosswork assurance is leveraging this routing components of CNC to provide some more advanced.40:36More advanced KPIs and more advanced view of the traffic traffic performance.40:56So just to explain on our side, on the cnc side, how we can leverage and enhance CNC routing analytics in CNC. The idea is that you can once again you can select two two nodes, one flex algo, display all the.41:02The CMP path.41:22Between those two nodes. And the idea is that on the CNC side we have some information about the latency, it could be the IGP delay today or it could be SRPM or something that we call IPM in the future when I mean when when it's when it's available, so with a combination of.41:23All this information, we provide a complete view of the of the of the of the path and the path stability and the path performance. And of course you can combine also with the bandwidth information so you have the full view of the of your of your path and and performance.41:44And of course you can have some some historical graphs if you want to see how the path evolved over over time. Let me pass the ball to Sujay maybe for the the 1st demo of.42:01Are you ready for the demo? For that any questions for Eric? Yeah, in the meantime, yeah, of course, if you have questions.42:24While while NO question? I was too fast or too. So there's one question Eric. IGP delays based on.42:34Yes, this is a very good question. I will need it should be I don't I will take it offline because I don't want to say something completely wrong. I would say yes.42:57But but probably we need to double click on this one, yeah. So I can. So the question is this IGP delay is based on yeah so in the in the IGP delay case, we are using the onboard router based capability.43:12Piece on the link where we start the performance measurement where it sends the probe packets. So yes, it is based on the T map, but it's only to collect for the metrics between the two, two routers that is between the head end and the tail end.43:32And then it is advertised into the IGP for all the links that's part of the ISIS team. Thank you. Thank you such.43:47Alright, if you're able to see my screen, I'll start this. So it's a two nodes. Yes, you can see your screen. Sure. So this is a demo for the path analytics where we have a topology today that we can visualize across different flex algos. Like this one is for flex algo one to.44:05For all the member nodes and you can start querying for path analytics, so as Eric mentioned that it's a capability of CNC today, you can start a subscription and register a path that you want to start the registrations for, which is to say between which.44:26Source router ID and the destination router ID and you want to get the ECMP paths, the stability of the ECMP paths for all those endpoints. And then you can also receive subscriptions. So it's a base URL subscription that you can start receiving it and it will start advertising this to your northbound system.44:45Now this is the mechanism today that we have for anyone using through APIs. Telling me that between R nine and R one in this case, what is the delay? What are my links that are part of the end to end paths and what is the delay and the link bandwidth as part.45:03Of those links and the paths for the end to end topology.45:23Now I can also, you know, create subscriptions, register a path, unregister a path if I'm not listening to those subscriptions, right, where I'm not monitoring in that case. But also at the same time, I can also delete for the subscription so that your OS system OSS system on the top can receive the subscriptions to perform for any use cases.45:29One of these use cases today on the CNC UI, we visualize it for a layer three VPN service or a layer two VPN service that uses locator as a steering mechanism, which is to say that in the service model context.45:49We have a layer three VPN service defined to tell it to use which locator across the network that I want the service to be provisioned. Now when the service is provisioned across a certain locator, since there is a baseline underlay config that's already there, the locator is attached to a flex algo.46:04The transport, I can visualize this on demand. When we say on demand is at any given point of time, I would like to know what is the path that this service takes on that particular flex algo that the service is part of. I will visualize the parts that the service takes. That is it is creating an ECMP tree if there are ECMP paths across the network and it will.46:24Which tells me what is the remote and the local interface for each of those.46:44Links and all and with IGP metrics that is collected across the PM on those links, we can see the link PM statistics also for that service. So this is the path analytics API today that we have this as a capability and you can then deploy the service for using your path subscription.46:49So there's a sample GRPC collection that's also available. So it's for both mechanisms out of the box CNC use case, but also to customize and develop more use cases as you require for your deployment.47:09So in the snapshot, we showed what is the response looking like, but then after you run, you do a get registration for all the parts that are registered and you can also delete the registration for any of the parts that you're done with monitoring on the use case of the service.47:24So this was a preview of the demo for, for part analytics. Back to you Eric. Thanks. Thank you, Sujay. Any question to Sujay before I go back to.47:41Okay, and I realized that we are we are using routing analytics or path analytics.48:03So I hope it's not confusing but sometime we we we switch from one terminology to the other so it's not maybe not easy but ok let me let me go to move forward if ok the system let me.48:11And start discussing about another observability tool which is called a distributed demand metrics and which is basically a way to offer traffic visibility and very quickly just after traffic visibility to do some capacity analysis through ACP, so.48:30So, so let me present what we, what we call DDM and ACP and then we will have another demo by, by Sujay around those two, two tools. So the idea of DDM is really to to to collect all those locator counters SRV six locator counters, and to build and knowing the topolo.48:50Building the the demand metrics and ACP, what we call ACP automated capacity planning will we will execute a few what if scenarios like a link failure, not failure.49:10Traffic growth and try to identify potential potential hotspots or potential failure scenarios. So DDM in in you will see that it's part of CNC, it's it's completely embedded in CNC and.49:24What we are trying to address back to the discussion we had around SRMPLS and SRV six use cases, is that.49:44In SRMPLS, we don't have this, we don't have an easy way to build the real time and accurate traffic metrics and it was, you can do it but it was complex and it was taking time. So here, and you will see when we go into the architecture, it's very.49:54Simple and and and very efficient.50:14And you have this full mesh end to end traffic traffic visibility and it's accurate, so we call it determinacy because there is NO sampling. It's real time accurate. It's if you are using Cisco 8000 routers, it's.50:16Building hardware. So, so so so so the accounting is based in our so it's it's very precise and efficient.50:36And the idea is really to, to have this this real time lightweight matrix, and that will open new use cases. I already mentioned capacity planning we discussed briefly about about bandwidth optimization.50:45Coordinated maintenance. You can, you can imagine a bunch of of new use cases coming out on top of this of this DDM it is supported, it's supported today across the XR, Cisco XR portfolio from the 8000 to the to the.51:02And this is something we are pushing to the IETF as well.51:22I did not check if it's still a draft. I think it's still a draft, but we are pushing hard to to get it to move it forward with ITF as well. So just a quick reminder and by the way, this is not this is not a CNC screenshot but a crosswalk planning screenshot because.51:27Because crosswork planning has been using demand metrics for for years and as I was mentioning, we were using things like netflow or to build it but but now the demand metrics, I mean this based on SRV six counters is much more efficient and much.51:46More accurate. That's what you we used to have to in the past. And the goal is really to, to know precisely where what are the traffic flows, where do from where to where we have we have traffic flowing and.52:06And that helps you to do your capacity planning to do failure simulation and so on and so forth. So back to DDM itself and in a very simple, very simple.52:26Way to present it imagine you have your locators like A B C D and what you will get is per flex algo per source and destination information of where the traffic is flowing, so like the yellow traffic going from A to D.52:44And since we know the topology between all the locators.53:03Will be able to build this this matrix in real time and it is automated, you're you don't need, you have nothing to do except enable those locator counters. It's real time. It's deterministic because we are, we, we are not sampling.53:08Or we we don't try to approximate.53:28What we get, we show all the traffic and it's lightweight because it's based on telemetry and even at the device level it's it's fairly lightweight. So, so very, very efficient and it, it, addresses an issue that we had have had for many years.53:31In a very simple way, very efficient way. So, if you want to see how it looks like, so we have the it supports summarization we are.53:51Also SR policy counters, so if we combine the SRV locator plus the SR policy counters where we have.54:11A real, a very precise view of of of where the traffic is going and and that's how we we build those this this matrix.54:20So just as a summary, we combine the telemetry, the services locator counters with the topology that we learn using BGPLS. We have this demand metrics builder if you want, and we'll have one metrics per per algo.54:37That will be available in in CNC. And as I mentioned earlier, once we have this matrix, it opens new use cases typically automated capacity planning that we are going to discuss in a in a minute bandwidth optimization that we will discuss in 2 min.54:59And then coordinated maintenance and probably much more as we move forward. But the two main use cases that.55:19We are going to build on top of the demand metrics are really automated capacity planning and and then with optimization to start with. You will see during the demo just that that is going to provide but just.55:26You should you you can see now the the demand metrics in your in your CNC we are in we we started in CNC seven.two so it's available today. We keep improving the demand metrics itself, how we, how we collect how we collect the counter, we added the summarization and and so on and so forth. So we are.55:44We're improving the demand metrics accuracy as we move forward and and of course you can find it in in CNC you can find for a given domain I know the screenshots are pretty small, but.56:03You can you can pick a link and and see all the all the all the other metrics for a given link, the ECMP split path the the traffic the the amount of traffic and so on and so forth. Yeah, per link demand is quite efficient, so.56:22For a given link, you can see all the traffic that go through that link.56:41And the source and destination and so on. So as I mentioned, one of the 1st use case on top of DDM, so DDM by itself is quite interesting because it shows exactly where, how the traffic is flowing through the network, but now.56:47We go one step further and say ok let's try to simulate some, some errors or some failures in the network and what the traffic, how the traffic is going to behave if you're So that's what we are we wanted to do with automated capacity planning ACP.57:03Maybe one day we'll find a better name for it, but for now we are still using this ACP acronym, and the challenge was.57:23Very simple. We wanted to, to give a way to identify traffic bottlenecks or hotspots in real time or to anticipate on potential potential bandwidth needs without maybe going through the the heavy heavy lifting.57:34Of using crosswork planning or another planning offline planning tool. So the idea I say was ok we have those DDM counters.57:54And then why why why don't we simulate those what if scenarios on top of those DDM counters and see how, how those counters are going to evolve? And typically we are just simulating very, I would say very simple scenarios like.58:05But very common like link not failure.58:23Traffic growth, things like that and then the ACP is going to to show us where could be the bottlenecks or where could be the hotspot. And that's the goal. The goal is really to to identify very quickly where there could be an issue and anticipate on this. Once again, the goal at the end is really to avoid.58:26To avoid your congestion situation, a bad situation if you want.58:47And so, so typically a very simple example is if we simulate a node failure, what will happen to the traffic? So we know that the traffic is going is is flowing through that node. What's going to happen if we are if we remove the node or if we, yeah.58:51Let's see if the node fails and where where where is the traffic going to be rerouted automatically and will that create congestion somewhere else in the network and where?59:10So that's typically the kind of question we try to to address with those with this solution. How does it work? Typically every 15 min, we get, we still need the routing topology, we still need the demand metrics, and then we are going to.59:24On those scenario on this ACP engine if you want.59:44Is going to run the scenarios and to provide the the outcome. And the idea is really to simulate link failure, not failure, traffic growth, SRLG failure, and to compute the KPIs for all the congested situation we we find in the network.59:48And you will see during the demo, then you can have access to the, to those KPIs in in in CNC. So typically you will see the total demands you will see the affected traffic when a bad situation happens.01:00:09Where there is excess traffic and what are the links? I don't know if it's big enough on your side but the links that are most likely to be congested or the ones that can be the most severely congested due to failure. So you know that those guys, those links are potential.01:00:29Not point of failure but point of, congestion that maybe you should you should and you should react I mean maybe you need to add more capacity or you need to add more D two or D two or or backup path. I would say that's that's more or less up to you to.01:00:49Decide what, what should be the right action.01:01:09And let me pass the ball to Sujay once again for a combined demo of demand metrics and ACP. In the meantime I can, is there any questions about DDM ACP?01:01:13So there's a question in the chat window. Yeah, so we are running flex algo 01:28 delay and normal isis routing in the in parallel based on SRMPLS. Is the simulation based on former former way and can it handle two overlaying.01:01:37Routing algorithms. I think the demand matri metrics cannot divide between FAO and FA 01:28.01:01:57I think we should be able once again I will let correct me if I'm wrong, but the demand metric should make the difference between flexalgo zero and flex algo 1128.01:02:07Yeah, to add to that, I mean if the demand matrix today is calculated based on the SRV six based locator counters and then from every head end to the tail end, wherever that flex algo path traverses in the network.01:02:22It picks the locator counters from the source to the destination and then detects for any incoming traffic all the traffic that's originated from the head end. So it's basically using the SRV six packet counters to show for that use case. So on the flex algo zero, if you.01:02:42You have attached it to a, to a locator and if that locator is assigned on on the specific head end, then I think as part of the SRV six locator counters will be able to collect that.01:03:01Thank you Sujay. Thank you. Any other follow up questions for the team here? So on this question, there was also the question whether the simulation was based on the former way.01:03:17Product or not. No, I would say NO, this is a specific algorithm but what you can do on the other hand is to feed way on, I mean now.01:03:36This is something you can do, I mean we can we can help you, but we plan to have it as a, as a native feature where you will be able to, to, to push or or to feed crosswork planning with DDM, so we'll be able to run the sophisticated scenario if you want or on.01:03:56Top of GDM.01:04:16Here we are just, I mean that's a simple, very simple failure scenario, not not as complete as what course planning can do. Does that answer the question? Okay.01:04:17Sujay over to you, sorry. All right, that's good. Thanks Eric. Yeah, I think, yeah, please feel free to ask questions. We'll we'll maybe do our best to address them. So.01:04:40Yeah, I think what Eric presented so far has been on all the evolution that has been happening with SRV six. In this demo, what I will show is how we can today leverage within the same controller and the network, the capabilities for.01:04:54SRV six, right? I have a network today that is defined across either in two IGP domains or a single IGP domain and I can see my network across the nodes and the links.01:05:14And similarly I can go back to traffic engineering to see what are my flex algo membered nodes. In this topology I have one flexalgo to focus and I'm saying that I have flexalgo 01:28 across these member nodes. Now.01:05:28The use cases now that we we have traffic as Eric provided the input that I want to start collecting the matrix. This is in real time, right? It's not approximation, it is deterministic. It's calculated from the counters, the traffic and making use of the view of the topology that is there.01:05:43I am overlaying this information to tell me at any given point of time how much is my total traffic or the demand of the network?01:06:03And in this case, I can see between both algo zero and algo 01:28 the traffic and between Peb to PEA, I can see what is the demand on the traffic and what is the path that this traffic takes for Algo 01:28.01:06:10So it is telling me at any given point of time or as of now when I search this, I can see what is the total demand of the traffic for flat algo 128I can also see here for algo zero in this case because it's based on the SRV six locator counters. And I can see for the algo zero it's using an ECMP path between both pebbs.01:06:25The NPA and the total traffic that is traversing across these two endpoints is 93.56 megs.01:06:45Now, this also gives me the visibility to dive deeper into to show me what is the demand on the specific link, because my link could be catering to multiple demands and I can see for every link what is the percentage share for those source and destinations.01:06:53That is in this case, it may say that between PEB and PA for a certain locator prefix, I might be doing hundred percent of that link or I might be doing 25 % of the link because I'm doing an ECMP across all the four paths that's available. So this kind of gives you that analysis into where automated capacity planning determines that.01:07:11At any given point of time with this demand of the traffic, if there is or if there were to be a failure on the link, what are the probable causes for those parts or the traffic that is traversing on that network?01:07:31So multiple yes, you can visualize, you can see the demands, you can see the ECMP paths across the network. And then you can also look into more granular level details of how the traffic is carried across the network.01:07:44So we'll also look into the snapshot of how the total use cases evolve between the various algos, various source, destination headends, determined deterministically through the input of topology and the headend.01:07:59Now, as we look into capacity planning, this is where it does a simulation. This is an on capacity planning, which is to say online, which is for every iteration that is run, in this case it is running for every hour. I will know what will be the analysis overview, what has been the state of my traffic?01:08:17Over the course of the time and the duration, how the demands have changed on the network. And then I can simultaneously determine the congested links. If the traffic were to flow, what would be the congested links at a 75th percentile or a fully occupied link?01:08:37And how much is the affected traffic during that iteration if that were to continue. And in that excess, if it excess of the link during failures, then it's the worst case scenarios that we would need to plan for capacity in the future if there were to be failures on.01:08:54To the network and there's an excess traffic that is planned.01:09:14To go over the capacity of the link because there are NO other backup or deterministic parts. This analysis also gives you into the granular level details for every scenario to point to where those congestions might happen. And this is kind of giving you the percentile graph, but on the simulation here it tells me the occurrences, like for which interface.01:09:18Which node and how many of these occurrences could occur at any given point of time. So for both congestion scenarios, which is for link and excess traffic.01:09:38So what is the worst case utilization for the current link and what would be the access traffic? And then a link and a node failure scenario, that is if a link were to fail or a node were to fail, what are the saturation points or limits in that case that affects the traffic when that snapshot of demand was.01:09:49Collected from the traffic, right? And it will also identify if there were failures or congestion occurrences, then what would be the affected traffic and how much percentage of the traffic might be affected in the very worst case scenario.01:10:09So it's a future anticipation to tell you that when these demands occur in your network, these are your failure points and these are the affected traffics that might work well. So it's a pretty advancement on these use cases because today it gives you that visibility, but that's the whole concept for the medium and capacity plan.01:10:22Thank you Sujay. Any questions guys before I move on? Okay, so, as I mean we we discussed we just discussed DDM and and ACP as an ACP as a 1st use case on top.01:10:43Of of DDM the.01:11:03The other use cases is around congestion mitigation. So we say, ok, if you do your planning properly, you don't need congestion mitigation but bad things can happen. So the idea is to you will see in SRV six either to do a local congestion mitigation like we do in in SRMPLS and that will be available in in 20.01:11:0627 with CNC, but we are also working on another more advanced solution where we are trying to leverage.01:11:26The DDM information and try to reroute the traffic in a more efficient way. So we assume that right now the has been done properly and then a bad situation happens. Since we are going to detect the potential congestion just like we we are we were doing with with.01:11:35The idea is really to try to reroute the traffic, not, not locally but potentially on a larger distance if you want. But by leveraging the traffic demand, we know exactly where the traffic was supposed to go and we will reroute the traffic according to the destinations.01:11:57So we are really going to leverage a demand awareness.01:12:17To compute the best alternate path if you want or the less disruptive we want to minimize how much traffic we route and and if possible we want to make sure that we are we take the most take the most efficient path. Alternate path, of course.01:12:21And once again we are going to to provision what we call tactical or ephemeral SRV six policies and we will see only the impacted impacted flows.01:12:42And the idea is we to mitigate what what we could not what we could not plan. So any unexpected unexpected situation. So it's a very, very simple overview of what we what we have in mind and what we are working on right now. This.01:12:57This is what we call global congestion mitigation that should be available hopefully in at the end of 2027 but if you have some interest into it, let let us know and we can discuss more, more in detail of course. Before I wrap up this session, I will.01:13:16Give the ball back to Sujay for a few a few slides on on the migration and how crosswork automation can help the migration to SRV six.01:13:36Do you want me to, do you want to share the slides or do you want me to to move for the for you? Sure Eric, I think if you want to continue, that's ok. I can talk on over the slides. Okay, ok, let me let me go back. Let me go ahead.01:13:48Over over to you as you want to do before we move on, there is a question in the chat asking how DDM was built. Is it measured.01:14:09At all head ends per locator and calculated for intermediate interfaces or is it measured at all points in the in the network?01:14:25Yeah, so I can take that. So today DDM is calculated on every head ends. I think in one of earlier slides of Eric, there's a demonstration that shows for every source to destination head end, whatever is the incoming traffic, the locator counter.01:14:39Users will.01:14:58Find that traffic, then there is also traffic that traverses through the network. So if you want to accurately measure, there is a statistical calculation from the traffic counters that is determined. And then we also calculate for the path that is in the network, so it calculates what is between the source and the destination for the.01:15:00Specific flex algo or the locator that is originated from here ok so do you want to cover the slide or do you want me? Yes.01:15:20Okay, sure. I'll go through. So, whatever we presented today is to show that the capabilities for the technology is required and the need and how with crosswork this.01:15:37Helps to simplify, gives you that visibility on a real time basis and deterministically, right? Nothing is approximations, but we also find that today a lot of customers are in the evolution journey towards the service and there are many questions, right? All of us would have that if I am migrating my network.01:15:54I'm migrating to a certain technology, what happens to all my legacy services, what happens to all my legacy deployments, how could I measure and monitor that it is accurately doing what the intended technology is? And how much can the controller do this? I mean, as for one of the questions I have both network opera.01:16:14Or I have MPLS and then I want to run both dual plane network topology. So because of this complexity, today we know that with the right set of tooling that Crosswalk offers today, the migration can help ease that job of someone doing manually.01:16:37For all these operations and have an automated tooling that can help drive this migration.01:16:54Can we go to the next slide Eric? So in this scenarios that we find that customers can use is, can it be automatable? Yes, it can be automatable through cross work automation. And we can show how this is done in the next slides.01:17:01So we find that the brownfield migration is predominantly into these four phases, right? Which is from phase 123, and four. So you 1st have your initial state, which is your MPLS underlay, your traffic, your overlay services are all sent over your MPLS traffic, you have.01:17:21The routing enabled. You can start enabling IPv six routing, like you can start introducing new route reflectors or SRV six, and then on your services, like on your PEs, your endpoints that are catering to services, you can advertise VPN services.01:17:40With both MPLS labels and SRV six, right? So it is a dual connected P solution. And through BGP policy measurements, you will determine what is the preferred SRV six advertisement with, you know, granularity. I mean, you can say.01:17:58Well, I want this service but I also want to pick a certain prefix. Yes, you can go to that granular level or you could tell for all endpoints under that VPN service. In phase three, you will advertise the VPNs with both MPLS and SRV six six and use the BGP policy to prefer which SRV six advertisements to use.01:18:14So once you start migrating to this SRV six service validation with pre and a post check, you will then end up in fully migrating your services and then removing MPLS from your network. Now, of course, when this looks on deployment, it's it's a phased approach, dual plane connectivity, and this has been proven that.01:18:33With the dual plane network, we can still have connectivity to both the planes and to services across the endpoints.01:18:53And can we go to the next slide? Sure. So what we use here is cnc's use cases, right? Now CNC today comes with an embedded workflow, which is a key capability that I can write my workflow at every point of time in this space. So as I.01:19:01Showed earlier for phase one to four.01:19:21I can have a pre check at phase one, I can have a post check at phase one, and before I operate on phase two, I can have a pre check on a set of operations and commands that I would like to run to make sure that all the required outputs that I expect are seen. And these are built through custom workflows. It's a workflow language environment. I can.01:19:23And build it through the workflow manager embedded into CNC because today the visibility of topologies there, the orchestration capability through NSO is there and with workflow, we can collect these data points from the network. Of course, this is, can be custom built but we have some early prototypes that we can share, but the workflow manager.01:19:43Really helps simplify this operation through for migration.01:20:03And can we go to the next slide? So in this migration, what we then determined is what is the workflow steps that you need. You enable IPv six addressing and routing through your workflow manager. Either you can say to use linked local addressing on all the interfaces that are part of your IP.01:20:08SIS. Second is enable SRV six locators. You can pick maybe two algos, algo zero and algo 01:28 or maybe with two different metrics that you want to use across your network.01:20:27And you introduce route reflector V six, right? So that RRV six and BGP V six sessions are introduced on those head ends that carry the services which need to learn from the route reflector is what is the best path or what are the paths that are being advertised to put it in its routing table?01:20:40And at step four, you are at the state where today you have a dual plane connectivity that is both MPLS and SRV six are still part of the network, but the VPN services still prefer MPLS underlay to provision to use as part of their best path for this.01:20:58Service. Now finally tweaking through the BGP policy local preferences, the BGP policy will prefer SRV six underlay over MPLS. This is the workflow that today we kind of have it as a prototype, but we have evolved this to also include how can you eventually remove MPLS in the state?01:21:18And finally with the adaptation in CNC because you have the workflow manager and the NSO integrated with workflows, you can also leverage the capability of what workflow manager has today and send out notifications because this is the key for any operator or an administrator to know what.01:21:38What is the state of my network if my migration has initiated and I want to be notified for every phase of the workflow that pushes this conflicts through this engine and to be notified that whether to proceed to the next step or if there's a failure is to determine on the notification that what is a failure response.01:21:58That was received at every step in that case.01:22:18So I know we are at the top of the hour, but we do have some previews and demos that we can share it later, but I hope this is useful for you all to see the value that Crosswalk holistically can bring together to help in the migration journey to SRV six.01:22:22Yes. Thanks Sujay. Any are there any questions for Sujay or Eric here? Okay let me maybe let me wrap up this this session and.01:22:43A few takeaways and and and some of the next steps that we have we have in mind. So 1st of all sorry excuse me. 1st of all a quick view on the roadmap, so as you know, it's a roadmap, so it may change with time it's not a commit, it's not.01:23:00Not commit, it's not a commitment except for CNC eight.zero.01:23:19On the left, which is more or less which is about to be to be to become available. And what you will find in CNC eight zero is mostly what we call the enhanced DDM or DDM phase two depending on who is writing, where typically we.01:23:24We we improve the scale and accuracy of DDM and we also summarization typically. And we are adding this SRV six policy counter collection and visualization to complete to complete the DDM in CNC eight.one that.01:23:45Should come in.01:24:04Hopefully in the 1st during the 1st half of 2027 we want to productize ACP. So the demo that gave you is still a demo. So if you want to try it in your network, we'll be more than happy to help you, but it's still at at a demo level if you want.01:24:05It's not really a product guys, so that's something we want to do in eight.one you should see also the 1st or.01:24:25I should not say the 1st but what we call the bandwidth observability dashboard. So, we have DDM, we have ACP. Now how we have we have topology information, how do we combine all this in a, in a, in a dashboard for observability? And you should see some the 1st bits for.01:24:34Basics aware local congestion mitigation.01:24:53Okay. And then we are going to add more more and more use cases around bandwidth management and we will introduce what we call a bandwidth manager. If you are very familiar with CNC and very familiar with SRMPLS style, safety style has already already a bandwidth manager function.01:24:56And we are going to enhance or provide a new bandwidth manager for for SRV six use cases in CNC A.one. And I mentioned it briefly earlier. We will have this DDM export tool crosswork planning, so if you want to use.01:25:16So DDM data in crosswork planning that will be that will be possible as well. It's possible today but that that will be fully fully supported. And then on the radar we have this global mitigation that I was describing very quickly earlier.01:25:36We should see some SRV six style capabilities, predictive analytics, and more and more agent use cases and recommendation around SRV six. And just to conclude the one last slide what we would like you to keep in mind.01:25:541st of all that all those SRV six innovation that makes SRV six automation and SRV six operation very easy and very efficient.01:26:13And and this and the key role of CNC in that picture where we combine or we connect service provisioning visibility, planning and congestion mitigation in one application. So.01:26:22For us, it becomes very easy to to to manage your SRV six network as just shown you, it's fairly easy to migrate to SRV six already with crosswalk tools. And once you migrated or once you set up your SRV six network, it becomes very easy.01:26:37To, to operate and to automate.01:26:57That's what we wanted to share with you today. Thanks again for your time and we hope it was useful for you guys. We can, we don't have too much time for questions but we can take some questions and we will follow up offline if needed anyway.01:27:01Thank you. Yes, we'll hold it for any last minute questions for Eric or today. Yes, there's one question from Marco. What do you see as a killer use case besides SRV six locators?01:27:19To go to the effort of network migration.01:27:38Oh, that's a philosophical question, I guess I think I think you you have SRV I mean, I think that SRV six by itself has a lot of benefits and, and.01:27:42Very interesting use cases where in here today we are just focusing on.01:28:01The automation part of it, but I think that SRV six when it comes to service chaining and and other very advanced use cases today is really the key the key technology. So I'm not sure I have a good answer for you.01:28:06For me it's more I mean you have to to to to look at SRV sixt I mean as a whole and there are there are a lot of benefits of switching to SRV six. Yeah, specifically if you're looking for some specific answers, I think it depends on the.01:28:24Is provided to service provider or customer to customer. At least some of the customers I spoke with, they found the unified data plane end to end all the way from the.01:28:44VNF to to the end station to end station that that seems to be a killer app. There are, like you said, the locator, the the counters the, then.01:28:57The, the metrics on the interfaces and so on and so forth is another killer app at least in my opinion. Awesome. Thank you very. But if if there are any further questions, we can take maybe.01:29:13One more but.01:29:33If not, thank you everybody for joining this session. This this this concludes this the three three topics on this series part to autonomous network series. We will have two more few more coming up in the next in the couple of months. You can.01:29:36Find the upcoming series on the link that I've posted on the chat window and the recording for this session will also be posted on the community page available posted again on the chat window.01:29:55And you can like we'll we'll we have more, and we have few other series scheduled in the in the coming days, right immediately after this is SRV six happening in the Americas and different series happening in the European region.01:30:07Please take advantage of all the expert inside series and to learn from our experts what they have to offered on the topics scheduled on for the sessions. With that we conclude the session for today and thank you Eric and Sujay for the great presentation and the demo, and thank you for.01:30:27The panelists for helping out and supporting the session today's session. We will stay in touch and if if there's anything that you'd like to know, please reach out to us through via email or through the support portal. Thank you so much everybody. Have a great day ahead.01:30:47