Transforming Service Provider Networks Through AI-Powered Connectivity

Uh, today we made a, you know, big announcement on the Agile Services networking, which is a new concept that we are introducing for the new AI era on the traffic growth and managing the service provider networks and beyond. So today we are gonna be speaking actually, like, you know, three different topics. We have three speakers in the room.

First of, first of all, I'm gonna be describing, like, you know, uh, why we are doing what we do on the Agile Services Network. And then I have two colleagues with me, um, uh, Phil, and uh, and Brian. They're gonna be describing like what is in it, and, you know, which customers, and you know, how the customers are implementing actually that solution.

Okay? So three agenda points. So let's start, you know, why now and what we do now, uh, in, in general, okay?

When we see the traffic and when we see the service provider evolution and the networking evolution, actually, like we see AI is really impacting, uh, how the traffic mix is, is changing in the network implementations, right? So as of today, like 30% of the traffic is driven by the ai, and we already see the evidences in the data center interconnect and the edge to network actually flowing, uh, increased data, data rates with, you know, 30, 35% in the networking. So that's really like a mix of traffic, which we are not seeing, which we were not seeing earlier in the service provider networks, which is actually figuring a new thought, like how we gonna address the growth, right?

In this one. So, by the introduction of like, you know, chat GPT in 2022, now in two years, we are seeing like, you know, a huge adoption, which is 65% of the, you know, the traffic has been, um, you know, produced in the edge, uh, data centers or in the edge of the locations, uh, let's say. And then like we are also seeing, like, you know, up to, you know, uh, 65% of the organization using AI for the generation of the traffic and the content, right?

Uh, that trend is expected to grow, and by 20 20, 20 30, we are expecting a huge growth, like, you know, 120%, uh, year of increase in the traffic and the traffic metrics, like, you know, interconnecting each sites, right? So like the big question becomes like, how are we gonna address that growth and how are what we gonna address that adoption, right? But you can see like three problems, uh, on the space.

Like, first of all, as we described, increased traffic on, you know, triggered by the ai, and that bandwidth cannot be accommodated actually in the networking layers if you have a suboptimal transport, which is in today's architectures, right? So that will result in redundant investments that will result in like, you know, a suboptimal SLAs in the service introduction. So, like, you know, that needs to be solved.

That's problem number one that we are solving with the agile services networking problem number two is that, like the most of the service providers needs to innovate and introduce new services so that they, they can generate like three revenue and they can generate actually margins. However, with today's, like, you know, constraints, scalability, and like, let's say, um, squeeze margins that they have, it's not easy to actually innovate, like using 20 years of architectures and 20 years of, you know, suboptimal products, right? Uh, in the network.

So that's the second piece that we are trying to solve with the agile services networking. And of course, like, you know, the, the changing in the traffic patterns, as I tried to explain earlier, uh, it's a different traffic meets. Now it's edge to edge, but edge is supposed to be anywhere by the, you know, by, by the customers and by whoever is using this one.

So, like, you know, we need to introduce the anchoring points into the networks where you can actually float the edge in as a, as a function from starting from the access to the aggregation to the central of the networks without, you know, having any compromise to interconnect any endpoint right to each other. So this is the third point that we're gonna address with the Agile services networking. And let me actually describe how we gonna do that.

Like, how are we gonna solve this problem, uh, by addressing like, you know, uh, with three key pillars, right? Our first pillar is actually, uh, intelligent aggregation in intelligent distribution. So what it really means is that we need to push the edge as much as possible to the end user who is consuming it, right?

It means like earlier handoff of the traffic to the possible clouds, to the possible endpoint. So pushing the, you know, the anchoring point as much as possible to, you know, closer to the end user, right? That's gonna give us like, you know, an efficient service, you know, delivery, and we can, we would be able to interconnect any endpoint without actually making a tromboning in the traffic.

So like east to west, and, you know, south to north will be optimized by the traffic deliveries as well. Um, this is gonna provide us a network centric, actually resiliency in the networking, because now, like we're gonna have maybe a smaller edges or smaller anchoring points for all the interconnectivity types, which is gonna, you know, come up with, you know, reduced blast radios in terms of failures and easy management of the traffic, you know, while we are, uh, interconnecting the, the, the sites, right? So that's the first point.

The second point I'm gonna discuss with the assured convergence, right? In order to achieve this, you know, agile services networking connectivity model, you need to have edge anywhere, first of all, right? So that's really a convergence of like traditionally known access aggregation and edge domains, all, you know, in, in a, in a conjunction, right?

Uh, the hard point is like when you push, push the edge with this, you know, current architectures of the products and you know, the services like to the access that becomes too painful to accommodate the cost levels, right? So you need to address in a different way, like how much scale that you would be accommodating and how much capacity you need to accommodate in these different locations, you know, selectively. So it's gonna come with, you know, multi-layer convergence in the systems and in the architectures, like first convergence layer, we will be talking about, you know, the convergence of the residential and the business services.

So we would like to have a common architecture, common product portfolio, and common delivery architecture for the automation. And, you know, the assured, you know, network operations to accommodate, you know, both residential and both business services altogether. The second piece is gonna come with the infrastructure convergence.

We are already doing this one, rather. Optical networking is a great example, but we are converging different, you know, technological layers in the networking, like IP optics, uh, passive optical networking, all in a common infrastructure, which we can actually do any service or any access point, any access type into the networking. The third point of the convergence, that's gonna come with wireline and wireless convergence, which is also gonna be addressed like, you know, the mobility services plus the, you know, the residential subscriber services.

So that common infrastructure will be serving to any access types for, you know, starting from the mobility and ending in, you know, the regular normal, you know, home users as well, right? So that's our, that's our second pillar, uh, to serve the networking, having a converge architecture and converge system, which we can serve any access. So how do we achieve this?

In the end is actually with the resilience simplification. That's, that's what we call, right? This is driven by the carbon silicon architecture, which means that across the portfolio for the pins that we are gonna deploy in access in a aggregation in edge and in core, we are gonna have a unified silicon portfolio, unified silicon one portfolio, which we can actually unify the services just by trading in and trading off like the scale and the capacity across the layers, right?

Which is a huge advantage. Like customers, when they need, uh, higher scale, they can actually trade this with the capacity, or they need higher capacity. They can trade this with, you know, the lower scale as well, right?

That's number one. The second piece is that the unification of the transport layer, Phil is gonna describe, like, you know, more details on this one, but that's gonna be a common protocol layer on the transport. So we will be unifying the service delivery architecture here.

Uh, the third piece is that we are innovating in the subscriber management side of things. That comes with the, you know, the functional desegregation as well. So like the, this new architecture will be served by the cos architecture on the BNG cos architecture from the mobility side as well.

So we are gonna have a common control plane, which is actually talking to our automation systems, and which are unifying the service delivery for subscriber management in the user plans area. And that's all gonna be connecting with the overlay business services, including sdwan as well, right? So these are the three key pillars, uh, which will be addressed within the agile services.

And now I would like to actually deliver the word to, uh, my friend, colleague, Phil, uh, to describe what is in it actually and how we do that. Yeah, kind of, what is it? Uh, so Agile services, networking, it's an architecture, it's a solution.

There's a lot of different components to it, uh, that make up what we call this end-to-end solution, end 10 architecture. And I'm just gonna touch upon a few of those today, can spend a whole kind of four hour session, going to a deep dive of that. But I'll really just cover the basics and kind of the basic blueprint of what it is, how it works, and, and how we're enabling, you know, connectivity.

I said AI is, is potentially a big traffic driver on networks, uh, in the future, but we're really covering any kind of operator service. So the idea is not just AI and the things that are coming, but all things that are, uh, you know, providers are doing today. And how we simplify those.

Again, it's about creating any, to any services. Uh, it's about taking away maybe silos of the network, where in the past we might have had an access network, a metro network, a core network, uh, and really being able to satisfy any service needs on at any point, no matter where it is in the network. Uh, the other idea is taking away potentially like, you know, platform limitations and having to use specific platforms, specific parts of the network.

It's really about being able to deliver those services across, uh, a single consistent, uh, set of hardware end-to-end across the network. Uh, so again, all these different service types over a single converged network. I think that's one of the key points Mustafa touched upon is convergence based on the network side.

And I'll talk a little bit about, we're also doing convergence on the, even the, the device side where we can do more within a single device, allow you to simplify the network, remove devices, remove protocols, uh, you know, over time. Uh, and it's really about simplification. I think something we've kind of touched upon a lot in the past is really network simplification.

That's really the key to, to networks that are easier to operate and easier to manage over time. Um, I know I just said the word simplification, but again, it's all about simplifying networks. Uh, so service optimization, uh, one of the key tenets, and I'll talk a little bit more about this, is like segment routing.

Uh, so obviously Cisco and segment routing kind of go hand in hand. They have for, for a number of years. Uh, but when you look at the actual kind of forwarding plane control plane of the networks that operators are building today, uh, any new network today is, is built using segment routing.

Uh, it's a way to simplify the network. You're removing protocols from the network, uh, you're enhancing, you know, services with the network, uh, and you're adding capabilities for things like traffic engineering. Uh, the ability to assure the network, and I'll talk a little bit about that, is all enhanced by using segment routing.

It's really the building block that we use to build these, I don't wanna say next generation networks. Uh, and it's really the foundation of what we're doing with agile services, networking from a controlled plane forwarding plane point of view. Uh, again, we talk about device convergence, and it's the ability to do more functions within that single device.

Uh, we've talked about routed optical networking now for almost five years, and that's, you know, the converge of IP and optical on a single device and eliminating a layer of the network and, and extra hardware that you no longer need. Uh, we're doing the same things with like pon, uh, so we have a routed pawn solution, very similar. We're able to do things like put an OLT within the device.

Now, I don't have to have that extra hardware sitting outside, uh, in the network. Um, and we're also doing a thing with those things with security. Uh, so I'll talk a little bit about DDoS.

So being able to distribute DDoS protection that any part of the network. So we're able to really do DDoS detection and mitigation on any router in the network, uh, which is a little bit different than the architectures that are currently deployed, where I might have centralized detection, I might have centralized scrubbing, and we'll move away from that and really, you know, distribute these functions, uh, throughout the network, and they can really be deployed anywhere in the network. Uh, I think that's one key point Mustafa made was, was distribution of, of services and distribution of, uh, these advanced, uh, functionalities over time.

Uh, we'll talk about Silicon one. Um, and these are all things that, uh, Silicon one is a big part of. I think the, the launch messaging and, and the keynote today, uh, is really extending what we've done with Silicon one over the last, you know, six or seven years, which is mostly focused on really high bandwidth, uh, applications and, uh, enhancing those chips, enhancing, you know, that line of hardware, uh, to serve more edge functionality.

So it's really now we have this kind of end-to-end unified story with Silicon One all the way from the axis, uh, to high speed core interfaces. So I said that's sort of a, a key thing that you'll, you'll see as, you know, new hardware comes out. And I'll, I'll talk a little bit about a couple of the devices that are, are coming out, uh, just now and in the near future that kind of satisfy those.

Um, we'll go through this whole slide, but really it is about taking, you know, what you saw at a lot of legacy networks, uh, the way they're built with, uh, you know, kind of the laundry list of protocols for doing things like MPLS label distribution. Uh, we eliminate a lot of that with segment routing. Uh, so we simplify the network segment routing just uses the IGP to distribute, uh, additional, you know, MPLS label information or SRB six and even simplify that even more.

Uh, and it's really about eliminating all those layers of the network, uh, and simplifying that. But then we take that to things like the sort of transition from IP to D-D-W-D-M, that's where routed optical networking comes into play. Uh, so we're eliminating a whole layer of this sort of optical, uh, electrical to optical transition, uh, and shrinking that into a plugable optic.

Uh, so a big part of the launch today was additional coherent optics that we're gonna use to enable longer reach Warner gig, as well as 800 gig, uh, just in a plugable optic that plugs into a, a, a whole set of routers that we offer. So again, it's about simplifying layers of the network, eliminating duplicate hardware, and again, doing the same thing with, uh, there's something we call private line emulation, which the traditional OTN networks that carry, uh, fiber channel OTN sonnet services, now we can do the same type of function, uh, within hardware on a router. Um, so it's really about enabling the network to carry, uh, you know, services, even non IP services.

Um, so we have that single converged IP network now they can carry, you know, services that are not just even ethernet and ip. They can carry fiber channel transparently over the network. So again, it's all about eliminating layers and simplifying that overall architecture.

I have a question on this slide. Sure. So, so, you know, there is plenty of like, you know, Cisco, uh, terms in there.

So, you know, how much is in terms of the interoperability between the different vendors, like, you know, there's obviously like on the left side, there are some, like, you know, uh, standards on the right side are these standards? Yeah, so when we talk about routed optical networking, it's pluggable coherent optics, and those are absolutely driven by standards. Okay.

Either OIF or open ZR plus. And we've been doing interop events now for, for years with that. And we're not the only ones offering from other vendors you might hear called coherent routing, something similar to that.

Uh, lots of interop there. Uh, like I said, it's really plugable, D-C-O-Z-R-Z plus optics, uh, uh, highly interoperable there. Uh, private line emulation.

Uh, there's some things we're doing there on the interop side right now. That's a pretty unique Cisco solution. Uh, you'll see at events like OFC that's coming up, uh, in a couple months, uh, you'll see that we'll be doing an interop there with private line em emulation as well.

Uh, and the, the standards that drive that, uh, there's some hardware stuff there that's a little more secret sauce from a circuit emulation, but all the control plane standards are all being driven through the IET app, and they're all multi-vendor. Okay. So, so certain features will be probably supported only if you will be Cisco to Cisco?

Yeah, some of 'em, like the private line initially, there's some innovation we have there that we just happen to, to come out with first. Uh, but we do see others that are looking at that. And again, the control plane of that is all standardized.

Okay. It's VPN based. It's, uh, you know, exactly, there's some, uh, additions we're making to it, but we're making sure that it's not just a Cisco only thing.

It's, uh, there's a whole ecosystem around it. Perfect. Yeah.

Thank you. So I'll talk just a little bit, uh, a few slides about individual kind of innovations. I touched upon these earlier, and Mustafa also touched upon them as well.

Uh, again, you know, Cisco 8,000 silicon one, and extending that family to, you know, to satisfy more use cases, uh, is a pretty key tenant of agile services networking. And that's something we'll just continue to build on over time with that, uh, that portfolio. Uh, and again, extending coherent optics, uh, I said routed optical networking has been very successful and, and collapsing those layers.

Uh, but being able to, you know, address more use cases with longer haul optics, uh, and also with 800 gig. So 800 gig is kind of the next evolution of those coherent optics and being able to carry, uh, do 800 gig single wavelength over, uh, you know, traditional DWM systems. Um, but then, you know, we talked about hardware and a lot of those things and convergence, but also the automation and the management software that goes with it.

Uh, so we talk about agile services networking. It's not just about the hardware, it's about the software that manages the network as well. Um, and that involves the last two bullet points there.

Um, Crosswork network automation, that's the, the brand name we have for our, you know, family of management software that manages, uh, these higher scale SP IP networks. Um, and you'll see provider connectivity of assurance, um, I'll say formally A CDN. So it's taking, you know, what they did in terms of you able to do monitoring and assurance, uh, with sensors at a very large scale, uh, and, you know, pipelining that data, uh, into their systems, collecting it, doing alerting.

Uh, but then obviously with Splunk, we have the ability to even, you know, do further integration between, you know, provider connectivity assurance and Splunk to, to add even more sort of, I'll call it, you know, business data into what we're doing with the data that's coming from PCA and from the network. Uh, and we really look at the network as a sensor. Uh, I won't talk too much about this today, but segment routing has a wealth of different things we can do to monitor paths in the network.

We can monitor latency and loss. Uh, we feed that upstream into tools like PCA, where that, you know, data can be aggregated, it can be viewed, it can be monitored, uh, and then we can also, you know, do additional alerting on that. Uh, and from there it can go into systems like Splunk or integrated into other systems.

So you could do, get more sort of business value out of that, uh, data that you're getting from the network. Uh, so it's really about being able to do the end to monitoring assurance, uh, for the new infrastructure that we're building and the sort of new applications that we're going to have to, So these assume that actually you would have the PAC competition element and collecting all the data and then filling into Splunk so that, that you can get some intelligence out of it. Yeah, potentially.

Yeah. Uh, like I said, that's where we probably would have put this sort of the PCA collector in front of that too, to get another level of aggregation and sort of data enrichment. Uh, and then yeah, that would go into Splunk, you know, and then potentially, like I said, be combined with, uh, sort of business data, uh, even more metadata.

Uh, so yeah, the idea is, yeah, yeah, that data, whether it's a PCE or some type of other sensor that's built into the network, all that data flows upstream. Um, yeah. Thank you.

And so, so is there any, you know, vision, what we've seen this morning are the keynotes with the DPU and stuff, like, is there any sort of a vision of Leo looking at that for the service provider world? Uh, I'll say we're definitely looking at it. Okay.

Uh, So we're not No commitment, but you're looking that, That was on the nexus side of things, but Yeah, Exactly. Yeah. Yeah.

But no, definitely looking at for, for applications that require, uh, more advanced packet handling mm-hmm. Things that just aren't really that doable today in high speed NPUs, but yeah, absolutely. Looking at Those 'cause they will be like, you know, in line what you were saying at the beginning, right?

Like streamlining everything you're doing end to end. So using same, same architecture, I, I guess, right? Yeah.

Yeah. Like I said, there's definitely a high touch packet handling functions that are, are better suited for A DPU than a network processing unit that we're definitely looking at those types of applications. It Does seem like a way to go though.

Yeah. Because anyway, the monitor then unifying the infrastructure in the case of route optical networking, they got rid of the device that was doing then the serialization between internet and the optical network. But once you get out of the way, then what are going, what are you gonna do to optimize packet handlings, serial ization delay, and all those things that you can beat with simple physics, you can, you have to get a bit more creative.

But the, the, but the moment you simplify it and it's in everything, is this in all the way from the beginning until the end of the network? And it does make sense that then it would apply all the power that Silicon War would have. Mm-hmm.

Because then if everybody has it, then what's the point of not making use of it? Yeah. But probably it's a bit early, but if it's a nexus, it will eventually land on some other product.

Yeah. So Nexus maybe leveraging that for advanced, you know, security, you know, purposes. And there's other sp things that we could potentially do with that, that DPU.

And again, like you said, it's about eliminating extra boxes in the network that have specialized functions and being able to combine those into a single, single device just to minimize footprint, minimize, you know, management of that, like A subscriber management, et cetera. Right? Yeah.

You using the DPU for that. Yeah. I mean, you could think of applications like nat, things like that, that, you know, modern high speed NPUs are not great at doing.

Mm-hmm. Uh, but, you know, those are applications to me. Those are things are things that we're, we're looking at sort of for the future and, and those types of applications.

Um, this is just an example of some of the newer devices that we're releasing. 6 terabits in a, a single fixed platform. Uh, but these new platforms are modular, they're a little bit lower bandwidth, uh, but they're meant for edge services.

So there's things like edge service scale that's, uh, quite a bit higher on these devices. Um, and then we'll have, you know, traditional, what we call access. Like I said, we don't really want to, you know, pigeonhole these as access and edge.

The idea is to support a common feature set across all of these devices. Um, so today you might pick a specific platform just because that platform, you know, uh, supports feature X, uh, another platform that's a higher speed platform, you know, doesn't support that feature. The idea is really to only pick a platform, I think Gustave said, based on the bandwidth needs you have, and potentially the port speeds you, uh, and not so much the chip and, you know, a specific device for a specific place in the network.

So again, you know, these are a couple of the new, uh, you know, systems. And the K 100 processor is the newer processor that we have on the silicon, one side focused on edge. The A 100 is a lower speed processor, but it's focused on lower speed applications that are usually kind of the, the access and, and very far aggregation parts of the network.

Uh, but again, it's a single architecture with silicon one end to end. And the idea is to be able to deliver these features, uh, ubiquitously across all of those different devices no matter where they're at the network. Uh, and I've got a couple slides just on like routed optical networking.

Uh, like I said, this isn't new. We, we talked about, I think, at Tech Field day over four or five years ago now. Uh, and it's really about taking what you see on the left is all these layers of the network potentially shrinking those down.

It's eliminating, and the, you can kind of see the simplification story there. Uh, traditionally in these, the networks, the way they were built, uh, the DW DM networks and things like that, you have a transponders, you have a lot of gray optics, you have to interconnect all of that. Uh, we've shrunk that all into a pluggable, you know, and that's, like I said, that was done, you know, five years ago.

And we're continuing to evolve that over time to, to increase the speeds and capabilities and lower the power that those optics. Uh, so when you look at Cisco devices today, and these are, you know, QSFE DD optics, uh, that are shipping today, but pretty soon, we'll also offer these in A-Q-S-F-E 2,800 gig form factor, which really unlocks a lot of applications for, uh, for more devices that the ability to use, uh, routed optical networking. Um, it's something just that just makes sense.

You know, today we have over 200 customers that have this deployed just because it's, uh, I call it kind of a no brainer technology. It's just something that you're eliminating, you know, devices in the network. You're eliminating complexity, uh, and it's, you know, cost effective.

And it's, uh, from a both CapEx and OPEX perspective. And, uh, Brian will talk a little bit about that. And some of the customers that we have, uh, have deployed this at larger scale today.

Uh, and again, the optics that we've used for routed optical to date, uh, they're 400 gig optics, but they support multiple speeds. But they do require like A-Q-S-F-P DD port, which limits some of the, the reachability and deployment to those, uh, those optics. Uh, so very soon we'll release a QS B 28, 100 gig zr, which is a tuneable DWDM, uh, you know, 100 gig optic that's in a QS B 28 form factor.

Uh, it's always been the, the limiting, you know, factor with that has always been power A qsb 28 ports were always meant to be very low power ports, about five watts. Uh, a 400 gig ZR optic is around, you know, 20 to 23 watts. So you can't really support that in a QSP 28 port.

Uh, but, you know, technology continues to advance and we've shrunk the power usage of the chips that drive the DWM, uh, to a, you know, within a 100 gig speed, uh, to something that's less than five watts. Um, so these are great for access rings, aggregation rings, things like that, uh, or even point to point dark fiber applications, uh, where now I have a tunable 100 gig, uh, SV 28 optic. Um, and then on the high end side, we're continuing to advance what we're doing in, in DD and 800 gig.

So 400 gig ULH takes the reach from a four oh gig, you know, coherent optic, which is about 1300 kilometers. And the current shipping, you know, sort of version of that to now something that exceeds 3000 kilometers. So some of the really long haul applications that we couldn't reach at 400 gig, you know, in the past few years, uh, those are gonna sort of go away now as we have ULH optics that have a, a much longer reach.

Uh, and those would be supported across all the devices that we, uh, support 400 gig with today. Um, and then 800 gig ZR ZR plus. Uh, we're starting to see, you know, more interest in that.

Uh, we recently did a couple of press releases with, uh, you know, two carriers in Europe that have tested 800 gig zr. Uh, and it's really about these high bandwidth, uh, interconnectivity things like DCI, uh, where we're addressing those needs with 800 ZR and 800 ZR plus. Obviously the reach is not gonna be quite as long with 800 gig, but you know, it's very high bandwidth applications.

Um, and those are things that are coming, you know, middle of next year or middle of this year, I guess 2025. And I said, we'll continue to evolve optics over time. And we're, you know, even now looking at the, the next generation of those, beyond the 800 gig, uh, I mentioned segment rounding is a foundation for how we kind of, if you take the layer above sort of the, the infrastructure piece of the network and how we address, you know, how packets are carried over the, the new infrastructure, uh, segment routing's, really the base of that.

Um, and people typically ask you, we have two flavors of forwarding planes with segment routing. It's really the same architecture with two different forwarding planes. There's S-R-M-P-L-S, which has now been around for quite some time, and it's pretty widely supported across the, the vendor community.

Uh, SRV six is the IPV six variant of that. Uh, and it really takes what we've done with SRM PS and it, it simplifies it even more. Uh, you're really just doing IP routing.

Uh, you get all the same capabilities you get with the MPLS, uh, forwarding planes, uh, just simplified with, with ip. Uh, we could do things like aggregation of space. Uh, you think of, of traditional, traditional I-P-M-P-L-S networks.

Uh, you have to have a loop back address, like a slash 32 that has to be reachable from every other part of the network. You can't really aggregate that very well. So you still have to distribute a lot of routing information end to end.

Uh, with SRV six, I can use aggregation. So I can have one part of the network that's all behind like a slash 48 aggregate route. That's, and I don't lose any of the same kind of TE capabilities.

So there's ways that we could still do, uh, you know, end-to-end traffic engineering. Uh, but a network that might have been, you know, 50,000, you know, B-G-P-L-U routes, I can now simplify to, you know, 40 I PV six, you know, routes. It's really as many regions as you have.

You can simplify the networks. So we've seen very large networks, uh, where if you look at the blueprint of how those are built, S RV six have very little routing information. And the less state, the less routes, the less scale you have to deal with.

It makes the networks much simpler to operate. Uh, you can use things like trace route if you, A VPN uh, service is really just an IPV six address. Uh, it's a constructed IPV six address, but you can trace route to your VPN service.

I don't have to look at a stack of labels for a service anymore. I can just simply look at a V six address and know exactly which service that, uh, that's going to. It also allows us to do things with, you know, network programmability that we weren't quite able to do with, uh, with MPLS just based on the scale that it, uh, uh, it gives us over time.

Um, so one thing we're doing, you know, with sr, and there's a lot of SR innovations, I just picked one to highlight today is something that we call integrated performance measurement. It's a combination of what things that we're doing with hardware as well as, you know, SR and SRB six, uh, software. But it's really, you can call it IP performance measurement.

This all just operates on the IP layer. Uh, but it gives us the ability to, you know, from a hardware node to measure every path across the entire network, uh, at high scale. And we report back on this, the, you know, the question i, I kind of posed earlier.

We've, you know, used the, really, the network as a, an assurance sensor. Uh, in this case, we can get the data from every path in the network between any two nodes on the network continuously. Uh, and to be able to report that into the tools that we have on the automation side, like PCA, uh, and like crosswork.

Um, so it gives you a lot of data in order to, you know, provide assurance to the whole network. This is monitoring the underlying infrastructure, but then I can add service layer data to that as well in order to, you know, correlate between the service layer of the network to the underlay part of the network. Uh, but in some of the newer hardware, we could set these probes as fast as, you know, 12 million, uh, probes per second.

I'm not sure you would do that in a production network. Uh, but it gives us ability to measure, uh, ECMP paths up to, you know, 128 ways if we wanted to continuously, uh, and again, you have to get the data out of the network, but the, the data is there in order for us to, you know, do more actionable, uh, you know, correlation and assurance than we really could ever do before. So again, that's just one innovation that we're doing with sr.

Uh, with, in addition, it's like measure the latency per link latency loss measurements, uh, within the network. And that all gets fed into external systems that we then use to correlate that data to the services that are carried over the network. Um, I think we mentioned a couple times around like device convergence.

Uh, we look at, you know, today we have kind of many different components that maybe sit, you know, outside the router, whether it's a transponder, a pon, O-L-T-O-T-N devices, D-D-V-D-M systems. And we talked a lot about router optical networking, and that's a pretty simple example of shrinking that into a pluggable. Uh, but again, with technologies like routed pon, we're able to do the same thing with, uh, an OLT that's doing XGS PON services.

So I might have a router that's deployed somewhere in the network that's doing business services, whether they're fiber based or, or some other type of, uh, business service aggregation. Uh, now I can also offer, you know, pawn endpoints off of that as well, uh, versus having to deploy a different OLT device, which is what's typically done today. Um, so that we call that kind of device convergence.

It's really being able to combine that. Uh, the other thing that we leverage more and more today is the ability to do application hosting on the router. Uh, so we can host third party applications and one that we're, you know, leveraging today does DDoS protection.

Uh, I mentioned earlier now that we can distribute DDoS protection and scrubbing really any router in the network. And we're doing that through that third party application hosting infra, you know, uh, infrastructure that we have that's, uh, specifically built within iOS xr. Uh, and it gives us the ability to not just, you know, converge the hardware on the network, but then also a lot of software functions as well.

Um, so again, it's that simplification. That's really the key part that we're, we're looking at with kind of agile services networking in order to simplify the network and be able to offer these, you know, different services everywhere across the network. And then as new services come up, it gives us the, you know, ability to even leverage, you know, applications that we haven't, you know, even looked at quite yet to, uh, for more advanced functionality.

One question for that. So totally makes sense to condense everything. To simplify, how is it with the telemetry from your perspective?

Do you still have differences in the quality of the measurements if you're going over an optical interface or an normal routed interface or the telemetry data? At the end of the day, you have the same quality no matter how the interface is looking like. Yeah, I think think the key part is to have like the telemetry.

Like if you're talking about the, like, so the hardware convergence part of it, uh, yeah, you have to have the same telemetry that you traditionally had in that other platform. You have to replicate that. And we've, you know, yeah.

Gone to great lengths to do that. Uh, so when, if you look at routed optical, like there's the thing, the, the telemetry that you typically go off like an optical transponder, uh, we have all the same telemetry that's on the router for that port on the router, uh, same for pond. So, you know, there's a, a wealth of pond data that's available.

Uh, and yeah, we have to replicate all of that same data on the router. I imagine if you let's say, uh, normalize all these telemetry data Yeah. Back in the days we had different islands.

Yeah. And then people needed to compare the different quality of telemetry data, if you unify, that also makes everything much easier to, let's say, scrap and then maybe send it to something like Splunk for additional analytics and so on. Yeah, I think that's really important, especially on the, in most networks that, you know, service provider networks are not single vendor, they're multi-vendor.

Uh, so you, like you said, that data normalization piece is pretty important when you look at multi-vendor and you do have to have some system that's gonna normalize that data. Um, ideally you have, you know, models and, and that you can use on the devices that are standardized. That's not always the case.

So yeah, we'd still do have to do some type of normalization at some layer to, in order to get into those higher layer systems. Yeah, absolutely. And we do try to use, you know, open models where we can, so we, when you look at things like the ZR ZR plus optics, the dcos and routed optical, we, you know, implement open config models that are pretty standard across the industry for that function.

Uh, and, you know, support those and try to support the latest versions that we can in order to, you know, make it less painful for, for operators to adjust the data and get it into a system that, you know, especially when they have multiple vendors involved. Um, yeah, I talked a little bit about routed pawn. Uh, quite simple.

It's taking, you know, a pluggable and OLT, which might have been in a chassis, uh, plugging that into a, an edge device. So again, there I might have, you know, EVPN type services that are for business that are on the same device. Uh, I can then, you know, leverage the same device for kind of multiple functions, right?

I said just makes sense versus having to put an OLT at a remote location or backhaul traffic, uh, via, you know, transport or some other means to a, a centralized location. And these are all components that are part of, you know, we call Agile services networking. Uh, like I said, there's, there's more of them we can talk about, but the idea is to have all these be part of a seamless solution.

Uh, we talk about, you know, secure DDoS edge protection, that's the distributed DDoS that I've, I've been talking about, uh, where we run these, you know, containerized DDoS functions within the devices. Uh, and we support a couple different use cases. One is traditional DDoS is the edge of the network where you're peering, or it's an internet transit connection.

Um, so we can put that secure DDoS edge protection container, uh, in that peering router and it, and do the scrubbing right on that router. So the detection happens there, there is a centralized controller that manages all of those functions. Um, it's quite easy to even deploy that centralized controller will deploy the DDoS container to the devices automatically.

Um, so it's not a lot of, you don't have to do a lot of CLI type of things in order to get the container and the application on the, the router. Uh, it's all managed centrally. Uh, so if I want to have that type of functionality at one point in the network, I can easily deploy that.

And then that, uh, device is ready as a, basically a scrubbing agent, uh, for the whole network. Um, and then also, you know, protecting internal parts of the network. Uh, so let's say I want to, you know, I have some customers that I don't really trust, so I'll put that type of function on a router that fronts that, uh, sort of residential part of the network, uh, that can protect, you know, uh, the rest of the world from my customers.

It's not just about traffic that's coming into your network, it's potentially putting what we call a kind of DDoS detection and mitigation all around the network. So it's both protecting, you know, your network from some of your users or the rest of the world from your users as well. Like I said, there's a, uh, quite a, an innovative thing that we've done with, with building it into the network and eliminating things like centralized scrubbing and centralized, uh, their detection.

I have one question to the DDoS protection. Are you doing this more or less based on sources or also on services that you say, okay, this customer has maybe here DDoS running this spot, is producing traffic on this certain, let's say port, we just blocked that? Or would it be the whole traffic for a certain source or destination?

Yeah, it could be, it could be granular even down to like a, uh, like a traffic payload. So like, uh, like is it, it's a traditional, like DDoS, it's a, an advanced DDoS sort of detection engine. It's, it's not just based on, you know, a volume of traffic.

There's signatures that are matched. So it does, I'm not gonna call it ai, but maybe machine learning type of activity that's being done there. So identify those attack flows and it'll deploy as granular of a rule as possible in order to block the traffic.

So it's not just by source destination ip. So like, you know, uh, it depends on the attack that's being, you know, carried out, but it's not just gonna black hole the traffic for that whole, that whole customer IP or IP block. It's gonna try to do it by source destination port, or even as, like I said, as granular as like a, a payload signature.

Yeah. So in best case, a customer is running a ransomware sending large amounts of DDoS to a certain services block, but you can still watch Netflix without any problems. Yeah.

Ideally if it's, you know, there's a, yeah, yeah. Ideally that would, that could happen. Okay.

Uh, you know, if you're not filling up all the internet, but yeah, that's the whole point is 'cause yeah, there's critical, like businesses get DDoS all the time. There's critical services that, you know, and it might just be an amplification attack that's just garbage traffic. Uh, but you have business critical traffic that they still want to get through.

Um, so the idea is, yeah, you don't wanna block all of that, just the part that you, I think in the past this was one of the major problems. If you block everything and sometimes, I dunno, you block the ip, let's say from a university and you're just not affecting the single user, you're affecting the whole business or so, yeah. Yeah.

And that's where we, we've, you know, providers really had yeah, more advanced sort of scrubbing infrastructure that did that. You're back calling all the traffic to a scrubber and then, you know, just feeling off the bad traffic there. And then you have to find a way to backhaul that back to this.

You know, this allows you to avoid kind of all of that sort of, uh, complex, you know, uh, clean and dirty type of networking you had to do on both sides of that, uh, that function. Um, I talked a little bit about automation. Uh, and again, you know, Crosswork Automation is our family of, of automation products.

We have something called crosswork Network Controller, uh, which is really what we use today as, uh, you know, an end-to-end management system for the IP layer of the network. So whether it's EMS functionality like, uh, monitoring alarms or software upgrades, but then it also has more advanced sort of TE capabilities, uh, and traffic optimization as well. So it's a platform.

And, you know, I, I could spend a whole session just talking about, uh, Crosswork, uh, but it's really for doing things like, you know, uh, ES function, uh, as well as, you know, service provisioning of the network, uh, within the family. We also have like CROSSWORK planning, which is a, you know, traditional IP network planning software. Uh, for those that are familiar with like Cisco way, that's sort of the evolution of Cisco way is, uh, a new platform called, uh, you know, Cisco or Crosswork planning.

Uh, but then we can also do management, you know, across both IP and optical as well. We talk about routed optical. Uh, we have to manage those end end, you know, DWM connections.

Uh, we use something called crosswork hierarchical controller to do that, uh, where we can manage both the IP layer of the network, uh, and those, uh, probably, you know, EDM services over that network, as well as the underlying sort of DWM uh, photonic infrastructure. So again, it's really about end-to-end automation across every part of the network is covered. Like you said, we don't, when we're come out with this new solution architecture, we don't wanna leave out how we manage, then monitor that network.

Uh, and crosswork is, is kind of how we're doing that. So Is Crosswork having on top of all their should and functions then it's also being able to work as you easy, but are all the functions that are included there? Yeah, like I said, I don't have a lot of time, but the PC is definitely part of it.

Uh, so we do have something called S-R-P-C-E, which is embedded into the network. So for, you know, basic PCE functionality, we, we use that component of it, uh, but it is a part of kind of the bigger Crosswork family. And then for some more advanced PCE functionality, like when we look at, uh, congestion management, uh, there is a more advanced PCE that's built into crosswork that does congestion management, and it can do things we call tactical traffic engineering mm-hmm.

To reroute traffic during congestion periods. And that we do use an advanced PCE. We also have something called circuit style segment routing.

Mm-hmm. Uh, which is a way that, you know, to maintain things like disjoint paths across the network, uh, bandwidth reservations. Another way of saying, sorry, another way of doing SR is just that you're not just having these unidirectional tunnels and you'll just do using the protocol extinctions, which is wonderful.

Right, right. Yeah. Yeah.

So it's just a big service having crosswork, and then you'll have several modules and each one of them is doing a different thing, but it's still, it's the same. Yeah. Yeah.

It's all under the family and the umbrella, but there's a lot of different functions within what we call crosswork network controller. Uh, and again, some of those PC advanced PCE traffic engineering functions, it's just another module within that, uh, that overall controller. Ah, wonderful.

Thank you. Uh, I talked a little bit about provider connectivity assurance, and like I said, call formerly a c CD N. Uh, and it's really the ability to, you know, bring in a lot of data, uh, you know, from different sources.

We talk a lot about network data. Uh, data can really come from anywhere. They have a pretty flexible collection, you know, infrastructure, uh, and the ability is to aggregate that data, get more visibility into the data, so there are, you know, some machine learning things that are being done at that layer of the network, and also do alerting based on that data.

Uh, and one, you know, nice thing with that platform is the ability to do things like create dashboards, uh, to be able to look at all the data in a, in a meaningful way for, you know, operations teams. Uh, that's probably, or even customer portals and building those types of things. So that's one of the key tenets of, you know, PCA or provider connectivity assurance, but then also being able to feed that data into other systems like Splunk, where you can then do, you know, more data correlation, uh, and more kind of, I'll call it, you know, metadata that you're adding to that data in order to get more, you know, business insights out of that.

Um, within this, like agile services networking, you know, there's a bunch of different use cases, A few, these are just sample use cases where we're building, you know, dashboards and advanced monitoring with PCA, and we're taking in, uh, in most of these use cases data from different, not just the network, but even the SD-WAN overlay. So in the SD-WAN use case, we're taking in data from the network infrastructure. We're taking in data from, you know, catalyst SD-WAN manager, and then we're able to do correlation between those two layers within PCA.

So I said that's sort of one of the key tenets, is to be able to take all this data from, say, disparate sources in the network, combine that together, analyze it and alert on it, uh, in order to be able to, you know, combine those views that typically providers didn't have before. Uh, and even when we look at routed pawn, it's not just looking at, say, a pond, OLT, it's also looking at the overall, uh, you know, pond landscape as well as the subscriber, uh, management piece of it. This is my last slide before I, I turn over to Brian real quick.

Uh, but what comes out of this, like, you know, we talk a lot, a lot about sort of, you know, agile services, networking, uh, what comes out of that, you know, and it's really a, we have different use cases that lead to kind of validated solution designs. We have our own labs where we set all of these things up, we validate them, and then we produce content like solution guides, uh, as well as end-to-end, uh, user guides and implementation guides. Um, so we don't just wanna talk about it, we wanna build a validated design, uh, and publish that.

And we've done that before with things like converge and transport, which is, uh, an architecture we've had in the past. Uh, so we really want to enable providers to have network blueprints so they can implement these, uh, in a repeatable way. Uh, and it doesn't mean you have to implement everything that we're talking about within the solution.

Uh, you can ob obviously take pieces of that and implement it in the network as needed. Uh, but again, it's a, it's an important part of what we're producing here is the ability for operators or se or people internally and externally to consume these, uh, different solution architectures, uh, and implement them in their networks. Uh, and with that, I'll turn it over to, uh, Brian, who talked a little bit about some customers that are kind of doing just that.

So I, I, I think you've seen here, you know, a lot of the capabilities, but it's good to put the reality on it. Now, do we have customers that are actually on this journey? And that's important to see.

It is a journey. This isn't something we thought of a few months ago. And, and so there, we do have a number of customers today that we're already been working with for a number of years and hitting their goals and actually as we go forward.

So if you look here, there's, there's a number of different ones. There's a lot of different areas, different segments, different parts of the world. But if we just briefly look at some of them, you've heard already the likes of bt, um, British Telecom, obviously based outta the uk.

And again, you'll see common objectives of what they want to get, right. You'll hear this word simple, you've heard it all over our presentation, you'll see it all over. We need to get simple, secure, and the word resilient.

They're really, really, and really what they've done with their global fabric here is really going and achieving that desire of network as a service or nas as a lot of people refer to it now. And again, the building blocks upon which you'll see today in our agile services networking. They are key to how they want to go around building network instances around doing multi-cloud connectivity, offering an SLA visibility, and again, connecting services to that.

And that's all part of what we're doing today. We're further enhancing how they can develop these services. The likes of Bell, you'll hear things as well around, you know, how do we get low latency?

How do we actually get things like iot connectivity? And obviously the road to SRV six and SR is, is very key for them. But very key here as well is, is removing the complexity, reducing the layers.

They talk about, uh, you know, transponders, taking transponders outta the network, and coherent optics. It's all about collapsing the layers, making it more simple. And again, you'll see some of these here, the likes of OTE.

It's, it's very interesting, obviously, part of the DT Dece Telecom group and emea. You know, we talk a lot here again about, you know, how we, we simplify down, but even simplifying down subscriber management, you'll hear announcements around kind of like control and user plan separation. So how, how do we get better availability, resiliency into the network, how we get better visibility, how you have a better use of IP resources.

So there's a lot of different operators and a lot of different segments here. But again, if we, if we actually look and we look specifically at a couple, you'll see Swisscom here, and Swisscom most people know is, is a tier one operator or telco in Switzerland, have entities in Italy now as well. But clearly they are, you know, a mobile provider, IP provider, residential and tv and had a number of different networks for, to, for all these different, uh, segments over the years.

But what they actually found is they needed to take that complex, fragmented network infrastructure and make it more simple. Actually at the same time, while making it simple, the word you'll hear over and over again is sustainability, power reduction. And definitely as we go into the world of AI now, power is becoming very, very important and in both their goals, but their objectives for scaling going forward, again, very important while they converged the networks.

So they took their parallel networks and you saw a reference to 20 different networks. I think that was clarified a lot yesterday with saying, really it was more 20 different platforms, converging them down into a single platform. But how do I make it simple?

How do I make it reliable and streamlined? And how they actually did that was common. Common hardware, common silicon.

One from end to end with a common platform with, you know, the optics come from end to end. But also we talked about resiliency and digital resiliency, which was really key for them. So you see even our customer experience, how they actually did their CICD continuous integration, continuous deployment of software and features as they change that network.

So again, really using the principles, what we see in Agile Services network, the innovations they've taken their network for forward and their opex in general has reduced by an estimated 40%. But as you said, the biggest gain they see and talk a lot about 90% reduction in power consumption with massive increases in availability and reliability right across the network. If we take another example of someone that's very much on the road and, and really kind of doing a lot of innovation in this area's cult, which is, is one of the global digital infrastructure providers.

Again, mass kind of reachability over 1300 data centers that they reach over 33,000 buildings, they connect. But these are probably one of the leaders, again, in this network as a service. And really what they achieve by that is how can they have cloud-like consumption of the network, how can have the visibility that we've never seen before.

And that's what they really prime, uh, a lot of their capabilities around, again, getting a lot of other advantages as well. So very, very clearly, they have very, very demanding customers. They, they, they really want to get kind of large scale.

How do we actually get to the, the simpler network operations, but also getting the balance between, you know, cost and scalability and simplicity at the same time. One of their key, key, um, areas here is that over 80% of all their connections, uh, in the core in EMEA are using Ron route optical networks today and increasing all the time. But it's the last two areas here that they cause a lot of differentiation.

Simplified automation, programmability, you know, on demand cloud-like that's where their differentiation. But you'll see here, and we talked about Cisco provider connectivity assurance. That is actually the tooling that gives them the assurance, but also the visibility.

You know, do I offer the service? Yes, I do. And here's the visible way to see.

So they have an absolutely cutting edge user dashboard that end users can see what services they are getting. So, you know, enhanced customer control in increased satisfaction. Very, very scalable, very cost efficient on demand.

And again, they're seeing 50% opex reduction, uh, in this area already, along with, as we said, 97% again, in, in energy, energy, energy use. Again, massive goal for them. And then finally, just taking another example of someone, Elia, and again, based in emea, um, you know, very much in, in the, but they are, you know, an internet backbone and, and they're, they're, they, the, the world's number one internet backbone, uh, at the moment.

But again, how do you get similar, um, kind of goals? And, and you'll see a commonality between all these, right? How do we make it more simple?

How do we make it cost optimized but highly performant? Um, but then getting the bandwidth to the levels that we need. And so a big driver here again, is the scale of silicon.

One, common, common hardware from end to end, you know, coherent optics and being able to go up to 800 gigs is a massive, massive driver over longer distances for them. Um, so again, the whole Ron networking is a big part, but you'll see again, the commonality here in the final two pieces again, is how do we get that visibility? How do we get the assurance I offer it?

Or how do I do the monitoring from end to end, but also how do I get that programmability? It's all about how do I get the consumption through APIs to a programmability of the network. So again, you know, very much the end result.

What we get here is, is very much a cost effective network architecture, a more simplified, easier to operate, um, from their point of view of management, but also very much, you know, reduced energy consumption and carbon footprint on their network from end to end. Again, very much back lined by, as you see here, 64% reduction in their CapEx by streamlining and simplifying the architecture. So I think again, we'll with that, we just, you've seen the guys, they've introduced what the industry is doing.

We've seen what Agile services network is delivering from Cisco. But also you see here many, many customer references that are actually now taking advantage of these innovations to actually deliver differentiation to their customer. Thank you.