Introducing Cisco C9000 Series Smart Switches

Thank you everyone for giving us this opportunity here. It's an exciting day today. Um, something that we have been working on for a while, and we are, we have just announced this morning in the keynote at Cisco Live.

Um, it's really interesting that, you know, in 2017, I was talking to some of you guys in a similar setting and we introduced the Catalyst 9,000, and that has become one of those products that has been consumed by hundreds of thousands of customers, super successful. And very shortly it has become a super comprehensive portfolio with a line of products, a huge power family. And, um, the amount of features and solutions that we have introduced on CAT 9,000 is just unmatched.

Well, what's happening in 2025 is slightly different than what was happening in 2017. And if you think about, uh, the workplaces that we have and the network that assist in the workplaces, the devices around us are changing. These were not the devices that we used to have around us in 2017, or even in 2020.

Some of these are very visible, like robotic arms and more, more of a human eye kind of, kind of, uh, devices starting to get into some of these environments. And some of these are very subtle, like your iPhone or your Samsung phone, um, with AI running on it. And these devices, although from hardware perspective, it is the same from intelligence perspective and the way they are behaving is different.

And so that's, that's a change. We have also seen, you know, one of the things, um, that we are seeing a lot these days I'm hearing from the customers is wifi seven is actually pushing the boundaries in real life beyond one gig. We have been talking about since the start of wifi six, six e theoretical limits, uh, or the theoretical performance.

But now if you plug in the port and in a normal environment, you will see more than a gig being pushed from the wifi seven, um, access points. And so, um, it is the, the new, really the new standard for wifi is a multi gabit. And that's something that we also have to take into consideration.

Another thing that's happening actually, uh, interestingly in this ai, and I'll talk about AI in a second, but quantum computing is becoming a lot more relevant, especially from the network and security perspective. Um, quantum computing is great because it is going to solve those problems that takes years today in minutes or seconds. Once quantum computing is consumable today, it's there, but it's not consumable enough for the general, uh, public.

Um, but once it is going to be consumable, it is going to solve a lot of problem. But guess what? The, what is going to also happen is the bad actors will start using it as well.

And so when the bad actors will start using quantum computing, they'll try to exploit some of these, um, uh, some of the data, some of the algorithms that we have from security and cryptographic, um, perspective, uh, with the help of quantum computing. And in fact, the problem is that threat of quantum computing actually starts right now because one of the threats that's actually happening today is harvest now decrypt later, which means that the bad actors know that in the next five years, I'll probably have my hands on quantum compute. So why don't I just go and grab that data that I cannot read today, but I'll break, I'll be able to break it in the next five years, right?

And that's a huge problem. A lot of public sector and financial institutions, you'll see anybody who is keeping the data for a longer period of time, they're concerned and there are new requirements and new compliances. And so that's something that we really need to look at when we look at the, uh, networks for the next generation.

What we are also seeing is in is a rapid evolution of ai. Like if you think about ai, I mean, AI started in the nineties, you guys remember some of those chess competitions. Um, but in the, what happened in the last two years has been just, you know, unprecedented, like in the last two and a half years, the amount of evolution that has happened in the AI field is just, um, is just mind boggling in some ways.

And so we started with generative ai, but we are going more towards agent tech AI agents sitting next to us, like G two mentioned this morning. I, there may be 7 billion people on the planet, but maybe there are 70 billion agents around us in future. Um, that changes a lot of things.

That changes. First of all, the amount of traffic that is being generated by all of these agents, right? And so if you, um, if you look at, um, and we have to do a lot of work here in this area, but if you start looking at the AI traffic today, in addition to the amount of traffic that is growing, the pattern of traffic is also changing.

And the pattern of traffic when you look at it and, uh, at the AI traffic today, there are a few things that are changing that makes us think what we have to do in the network. And one of the things is the rewards flow of the traffic, right? com going up is just a couple of kbs, right?

Kilobytes of, um, data that's going up. But then you get the entire website down, which is several megs in some cases, right? And so that's the flow.

The flow has been downlink more traffic, uplink very little traffic, but with the ai, the traffic is going in the reverse direction. Why? Because all these modalities of AI that we have that is pushing the traffic, a lot of traffic up.

So my wife talks to Chad, GPT, that's why it's going up, right? Uh, we are uploading a lot of, um, videos and, and, and pictures to the a to the, to the AI applications. And that's going the other way, right?

And that's a change, right? And so we have to take care of that in the network. The network which has limited bandwidth are showing up more already.

Uh, the network ha that has ample traffic, uh, bandwidth, that network will show some of these things in the future. Um, the other thing is these are burst traffic, right? Um, but it's slightly different than burst traffic is also LA latency sensitive traffic, uh, as well.

So a deeper topic to go into, but to address all of these challenges that we have talked about, whether it's quantum, whether it's ai, whether it's new types of endpoints and, and applications around us. We have to look at the, the networks and how we are going to address it in the next decade, right? And so that's where we are starting, uh, a new, uh, Cisco switching series, and we introduce it today at Cisco live in the keynote, and it is called Cisco C 9,000 Smart Switches.

Okay? Um, so we are starting with two devices. Number one, the Cisco C 93 50.

These are these stackable switches, and my team here is going to be talking about it a lot more, and we will go into it deeper. This is our next generation of stackables, and it's sitting right here. Um, and, um, you can probably see it at, if you are here at Cisco Live, you can see it as one of solutions.

It's, it has many industry. First, you'll learn more about it later on. I'll keep it short here.

But the second thing that we have introduced is, um, is called Cisco C 96 10. Those of you who remember the Catalyst 65 0 9 and Catalyst 65 13, or some of you guys have put Nexus 7,700 in your, in your campus score. This is the next generation for that, right?

These devices, all these three that I mentioned, they are coming very soon. End of support. Last year of support is coming in October.

Um, we are here to help you guys transition into this as soon as possible. Uh, but this is a 10 slot chassis, two for supervisor, eight for line cards, 25, 448 ports of 25 gig, and 2056 ports of a hundred gig, um, that you can have on this one. And, and guess what?

The, the back plane or the capacity of this box can take it all the way to 400 gig ports on all of these line card slots. So super rich from performance perspective, but it's also rich from some of the other things from security perspective that we'll talk about. Yes, Christian, 400 gig when it is going to be launched.

So today we are going to have very limited ports, um, depending on the architecture that depending on the, on the, um, configuration of your ports, it can very well be non-blocking. So the amount of tra amount of bandwidth that we have total is non-blocking. However, you have to look at how the asics are mapped and, and Kenny and Sean will talk about it a little bit.

Um, and based on that, there are going to be some scenarios where it going, it might be, it might have a little bit of over subscription. Okay? So when we are introducing these things, I wanna keep it very quick.

We are also making sure that we are drawing a line and moving forward and, and introducing a unified way of managing these boxes, right? Something that we have not done in the past. In the past we have had, you know, Meraki product lines for cloud management and then Catalyst product lines with Catalyst Center.

But going forward, we are giving our customers a choice to go and select the right device that they need, the right hardware that they need for their requirements, and they will have the option to go manage it with the dashboard with A CLI or API or with Catalyst Center. That's customer choice, right? And that is, that is something that is going to be very new starting with So, so Ben story here with, uh, red Eye.

So I assume there's thick walls between those three three, we're not gonna be able to, you know, be on Catalyst Center or Meraki dashboard and still access the console, like we've, so is is it traditional Meraki or is this traditional catalyst? Uh, very good question. Those lines are blurring.

Okay. Okay. And we are trying to make, with, with the new one, you'll be able to get to the CLI while you also have access to the dashboard.

That's, so will you be allowed to make changes when it's offline to fix things and have it sync back up to the cloud? Or is it, is it still gonna be read only from the console port? Uh, it's, it's not read only from the console port.

And that's the change, like in, in one of the previous modes that we have introduced in the past is, was that once you are, first of all, you had to, um, wipe out and reload the box in a Meraki mode. Yep. We are making all of that process very, very simple and some of it is going to be discussed a little bit later on.

Okay. One of the other things that I wanted to show you, I know you guys will love this. Um, the asic that we, that we have, uh, on both of these devices, the 96 10 and the 93 50 are both silicon one, right?

There are two types of silicon one, uh, uh, chips that we are using. This big one here, I'll pass it on in the, in the room here. This is K 100.

4 terabytes per second. On the 96 10, there are four of these. Okay?

And so that's why you see the supervisor is a two RU supervisor, um, because we need to cool it down, right? It needs some air on it. Okay?

So I'll pass it on, can feel it's heavy. Um, and then we have the second one, which is the A 100 that is actually here on this box that you see 93 50, and I'll pass it on here. You guys gonna have a quick look as well.

Um, These are the same chips then that, uh, you're also using in, um, 8,000 iOS XR routers. Correct? Some of the eight thousands are using this.

Yes. Yeah. Yeah.

And, and Sean is gonna talk about, you know, silicon in a little bit more detail, okay? And then it comes with, with iOS xc, and I'll not go into too much of this because, uh, Sean is, is planning to cover this, uh, as well, but iOS XE is being enhanced as it's bringing microservices, um, process modularity in the iOS XE system. It is also going to become quantum safe, right?

So QPC compliance is something that we are going to bring at the operating system level as well, right? Um, and then we have, you know, application hosting that is going to be enhanced compared to the first version of the CAT nine case that we, that we introduced. So this is going to be a lot more capable, you'll be able to run multiple applications, um, at the same time.

Um, okay, I'll skip this just in the interest of time. Um, when we introduced the, the Cisco SMART Switches, what we kept in mind is we need to have a hardware that is going to serve our customers for the next decade, right? That's for the AI era, and that's the hardware that we are bringing.

And that's, this is just a start. The second thing is security fused in the network, which means that all of these boxes are PQC compliant, right? For p qc, it's alar larger topic, but I'll say it very briefly, PQC has multiple levels that you need to achieve.

Some of it is in hardware, some of it is in software. These boxes have a chip that is, that is going to make it a PQC compliant. Um, we are, these, these boxes are also hyper shield ready.

Uh, again, this is readiness not hyper shield functionality. And then you guys have heard briefly compensating controls due to talking about it in keynote today, and that is on the roadmap. That is something that is, that is going to help with the vulnerabilities, um, fix quickly instead of waiting and or upgrading.

Like we have brought heart patches in the past, but right now we are with compositing controls. It'll be a matter of minutes to, uh, uh, to block it. And then finally, operational simplicity with platform flexibility, like I talked about it.

So we are giving our customers a choice to manage all of these devices with the platform that they like. With that, I'll hand it over to Kenny to go deeper into 93 50. There you go.

Yes. Is it just the 96 10 or do you have, um, a chassis that are small that have fewer blades in it? So, um, we, we already have a very comprehensive Catalyst 9,000, so that continues with the new smart switches at the moment.

It is just the bigger chass that we are introducing. Yeah, Nine, Seven on the roadmap. You just created a new product here.

Um, Good job. It's probably either gonna be the 96 0 6 or the 96 0 8, but yes, You are giving me ideas. Thank you.

Excellent. Now speak speaking of size or are the 93 fifties the same length as the 93 hundreds, or are we gonna be replacing all of our wall wall mount racks again to, yeah, So, so again, Kenya is gonna be talking about it. There are a few differences.

First of all, this is 18 ru, um, depending on which one you're talking about, 65 0 9, 65, 13 77 next to 77. These are, these have different sizes, but one of the important thing is this box is front to back airflow, okay? Which aligns really well with the, with where this book needs to go today, right?

Versus 6,500 that used to go in the wine closet a lot, right? So, um, so there are differences and we have to kind of take care of some of those. Thanks Mohammed.

Um, my name is Ken Lee. Very quick introduction. So I've been working with the, um, switching for a good amount of time.

So I actually, um, when we introduced a Catalyst 9,000, I was initially working on that team and subsequently now we're introducing the, um, Ciscos 9,000 and, okay, so Mohammad talk about that. There are two model that we introduced, right, the 93 50 and the 96 10. So I'll go into the detail specifically a lot of the hardware related and then Sean will go into more details after, uh, my 25, 30 minutes later and he will go into the ASIC and the IRS xe.

What are the architecture differences in that? Okay, so yeah, so the first one Mohammed mentioned this is the 93 50, sorry, what are the enhancements is in that the first, if we look at it, he talk about the, um, silicon one, asics, and I heard some of you asking about, is this the asics being used on the other platforms? Yes.

0 network processing unit within the Silicon one family. So there is a lot of enhance business in there that's specifically for a campus environment, right? The hardware net flow, um, a much larger, uh, table size for the ACOs and so on, right?

3 tail bits, right? Obviously if you look at the, um, the 9,300, you look at all the fun panel ports, you wouldn't be able to using all of this bandwidth. So your question might be, okay, why is this so big?

Um, high performance, the fund panel has using some bandwidth, but there's also a stacking as well. So PVS generation will have 9,300 x up to one terabyte of the stacking bandwidth. 6.

Okay? 6 terabytes for the stacking bandwidth. Okay?

Scale wise we talk about it. The e um, comparing to the existing 93 hundreds is four to six times higher in term of the, um, ACOs in term of the routing tables. And Sean will actually talk more about how we were able to get the ACL scale to be much higher, at the same time maintain the same power footprints and so on.

Okay? Um, the stacking in here, Mohammed also mentioned we changing the stacking architecture, but the interface to the user remain the same, right? Previously we use our own proprietary for the stacking protocol, but now we using the standard ethernet based protocol, which is EB, excellent, okay?

So that opened up a lot of more possibilities. So let's leave it over here. So in the future, we might be able to write on top of this technology and gave you more capabilities.

Um, mo touch on the securities, I will have a little bit more touch on the security later on. Okay? So that's from the front, right?

In the back there are also additional enhancement that we have done. Okay? 6, but even in the stacking power previously, I think if, you know, you might have heard some of the customers saying, okay, stacking power is great, but there is, you know, a limit of how much power we can send from one stack to the other stack.

We have enhances this, right? So now up to 55 amps, we can send it over to the other devices. High availability is a key in one.

Any of those networking elements, right? The first one is, now we have three fan modules is in here and they acting as uh, n plus one, right? Any one of the fan fail is fine.

Previously all of them has two power supplies, right? Either you can run it and combine a redundant, but now we have three power supply, which is much more modular. You can have a n plus one redundancy.

One of them is acting as a backup, was the other one. Okay? So here's the architecture of it.

Uh, the only thing that I wanted to point to you is in some of the models you might have multiple asics, right? 3 terabits, it's about 500 gig is for the front panel port, and the rest of them is using for the stacking interface. So in some models you might require two asics.

So within that two asics there are interconnection between the asics and there's also additional bandwidth in the back there to take care of the stacking. If you want to stack additional devices in there. On the right hand side is a very simple, this is the control plane.

And on your left hand side, this is the fore winding plane. Am I going too fast? We're good.

Alright. Um, on the stacking I mentioned a couple of things, right? 6 T and also using the standard ethernet, but there are additional things that we have enhanced with these technologies.

Okay? I'll go, I used an example to go through it previously. When you do a stacking, you have to go from, um, switch number one, go to switch number two, and then connection from two to three, four to five, and so on.

And then from the very last one, go back to the first one, right? This was a only way to connect a stacking in the existing 93 hundreds. But with the new standard, using the e VXLAN as the standard protocol and also using the shortest path for routing the traffic within between the stacks, we are able to provide you additional flexibility.

You can connect the stack any way you want it, right? Every switch have two ports in the stacking you, as long as those two ports are connected, you can connect it any way you want it. Right?

The the middle column gave you an examples like, I don't need a very long cables, right? And most only need to go to three switches. Okay?

And the last one is like, oh, I can connect it any way you want it. Now also, previously in a failure case scenario, if one of the stacking cable failed previously, if this one fell, all of the switch stacking bandwidth now is half. Okay?

6. Do you have recommended stacking patterns? Depending on the number of switches you have because like that architecture on the right, that makes sense for four switches.

Yeah. When you have six, you could replicate that down one more. Um, do you wanna say like, and I'm sure there's a maximum number on the stack.

I'm eight or nine, probably don't eight Today, and we, we are looking to increase it to more. Okay. We can talk about whether that's a good idea or not.

Yeah. But Uh, that's, that's exactly the same thing, right? If you reach about area domain, then you might not wanna go there.

Yeah. But in some cases customer do see a benefit like, hey, I just need to add one more switch. Do I build two stack or can I do it with one stack?

I I agree with you. We can talk, You do, we should talk to those customers too. But I'm just wondering that now that there's variability in the way that you, you deploy these stack the backend mm-hmm.

Is, is Cisco providing recommended models depending on the number of switches or recommended connection architectures? Because like now all of a sudden there's variability to where there wasn't variability before and I understand the value not having the bandwidth on the back mm-hmm. Between the connections.

Makes sense. Um, having it just be connected however you want Yeah. Is not great guidance to customers about how to do this.

Okay. There's there's gonna be better ways, you know, like, like that model on the right's. Great.

Yeah. One of the ones you showed before was awful. I wouldn't do it that way, right?

So, um mm-hmm. I if you don't, I just, maybe something to think about is to provide guidance about how to connect those stacks on the back plane depending on the number of switches you have. Absolutely.

We've certainly, I can look into it and because also depends on which module you have uplinks and so on, right? Right. We need to take those into consideration.

Maybe there is a, some best practice that we can provide instead of saying that, hey, you can do it any way you want it. Yes, you can do it any way you want it, but here is maybe the recommendation that we can give you them, right? This One on the right is with purpose.

Yeah. Like, like it's, that's not accidental. The way that those are connected, it's done for, uh, resiliency as those cables may or may not fail going forward.

Yeah. Like it's done with purpose. Some of those ones you showed before, sure they may work, but that wasn't done with purpose.

Like there's, there's reasons to do it a particular way. Absolutely. Question Josh from diversified, do you have any pictures or diagrams that all show stacking horizontally?

Because that may also come into play. Um, that is today. Today.

No, it's not the horizon. That's fine. Yeah.

This is all what good thing stacking so much is smearing, You know, when we talk about like making two different stacks, make a second stack And then maybe you'll be able to connect the two stack together, right? One. Okay.

Sorry, just one other real quick question on that. With the second, a, uh, the architecture on the right mm-hmm. That kind of implies that you could have up to two cable failures and still, still survive.

Well have, If this one failed, let's say assume would split. It would split. Yeah.

This one failed, then the middle one is gone, right? That's it. But then yeah, but the other, the rest of the three will still will be working fine.

And then yeah. Any other questions? Good discussion.

Good feedback by the way too. Okay. Um, Mohammed, touch on the, um, post quantum cryptography compliance, right?

Um, the switch that you that on this nine case switch today, um, starting with the uh, trust anchor module or the 10 module within in there there is a sudi, right? A secure, unique, uh, identifier that sudi today is, um, PQC compliance, right? I put a little notes on the bottom in here.

That is what we use to sign for those. Subsequently, once the device start booting up, right there is a lot of the certificates is in there because we have to check for the certificate for the os, the image signing, the boot loaders and so on and so forth. And then also on the, if we have a, um, we are gonna support the MAX act in um, IPSec for max on IPSec.

There's a two way to set that up, right? You can use a certificate or you can use a p share strings, right? With the certificate that also need to be PQC compliance.

And when we introduce those features, those will be available within iOS or succeed, right? It's not gonna be just, uh, specific for this specific hardware, but from the hardware level within the 10 module, we have that built in. So the foundations is there.

So any next layer you add that up to those will be added as well. And Mohamed talk about the compensating control, right? You heard from the keynote today, the actual product name is called Cisco Life Protect.

I will stretch, I give a little room between the Cisco Life people might say Cisco life and then protect is Cisco Life Protect. And that is exactly what it is. And um, from G two you heard that it is already available in the um, nexus os, right?

This compensating control or Cisco Life Protect is on a Linux kernel level. So once we have available in one of the os, those could be easily put it into the other like iOS, Xes or the routing or uh, iOS xs. And all of this will be easy going through, uh, the la the next slide, the last slide on the 9,300 is the application hosting, right?

For application hosting, you need a CPU, you need a dram, you need the storage, you need the, um, data interfaces. Would the 93 93 50 we have enhanced the CPU. We have used a faster memory and we enhanced the number of CPU you can use as in there.

And the KR port, which we saw the application port where you can actually suck in the data so that this container can be processed as data. All of those get enhances in here. And this is the foundation where we'll be able to hold all of this next level of the security application, the next level of anything that is bus work gen, AI related.

Okay? Alright. 96 10.

Okay. So all the, the security feature on all of this that I talk about on the 9,300 are available on the 96%. Okay?

So this is the chassis, uh, Mohammed mentioned this is a 10 slot chassis with two supervisors in the middle. Um, a slots on the top and bottom. 2 terabit per second.

Okay? 4 ts per slot. So a question was asking, well, can you do the 400 gig line, uh, line rates, right?

The chassis absolute is no problem. Now we have to talk about a supervisor. When I go to the supervisor, I go through some of those, uh, key components, okay?

The two versions of supervisors. One is what we call the SUP three. The other one is SUP three Excel.

The SUP three is based on the K 100 ASIC and the sup, sorry, SUP three is E 100, ASIC sup three XL is K 100. The difference between the two are scale and as well as the buffer memories, but the performance, the throughput is the same. So either one you're using, you get the same performance and some of those skills is in here, you can see is 2 million routes and so on.

Okay. Sean probably will go through a bit more details on that as well. And with this chassis is a, uh, centralized architecture, just like what we have with, uh, the existing modular platforms on the, um, catalyst family.

So we are introducing two new line cuts. So those will be a native line cut with the, um, this new chassis, but all your existing 96 0 6 line cut will be compatible in here with an adopter. So they're compatible.

But if, if we have a mixed stack of your new cards, your new supervisors and your old cards, do the old cards reduce the performance of the overall uh, platform? No, it would not. Okay.

Yeah. So this is the beauty of the central centralized architecture. The supervisor's the one who can show what among the bandwidth to each one of the slots, right?

2 T, so be it, okay? Um, the fan, uh, also in the fan tray, so, uh, Moham also mentioned this is a front to back airflow. There are four fan traces in factor.

They are acting as n plus one redundant, right? Any one at fan fail, the rest will continue to operate. Okay?

This is just a cool pictures, like how do you do it with front back air flows, right? You need, you need actually open space in your back plane so that the air can actually flow from the front back. This is using a industrial first centralized cable back plane.

Those are not traces, those are actually physical cable. So when they bundle together, you have a plenty of space. The air can flow naturally from the front to the back, okay?

And those cables, because they have a very high signal integrity, so they can carry a very high speed signal. So a hundred gig SOEs in here is no problem at all. Okay?

Centralized architectures just kind of reemphasize what we just said before, right? The everything's done in a supervisor, the line card, it just grab the data and send it to the supervisor. So the line card doesn't do any processing, it doesn't really matter.

I send one gig versus I send 10 gig. It's all, it's good. Okay?

The two supervisors, one is the E 100, the other one is K 100. You look at the characteristics it's in here, right? The performance is the same.

The difference in term of the scale one, the Excel has a much higher scale. Similar, if you are using the 6,500, remember we have a sub two T Excel and then normal sub two T, right? For performance, I say, okay, this is to the question you were asking on the performance SUP two, I mean, sorry, SUP three and SUP three.

Excel does not take a full advantage of the chassis. 6. Okay?

6, that means 32 port of a hundred giger line rate for each one of those A slots. Okay? Your question is, is this non-blocking or full full line rate?

It depends on the traffic flow. So each one of those silicon, one asic or four of them, right? 2 T.

So now the case is if all my traffic have the go across here, then the worst case is three, uh, two to one over sub supply. But you might not traffic flow in here, does not have necessarily have the go out to the other asics, right? They could also be going in here and coming out here.

And the way those asic mapping to the, uh, fund panel port is this asic, the first ASIC connected to all one quarter of the fund panel port on every single slot. And the second one is the second one, fourth, and then the third is third one fourth. So even if I have a single line cards is in there, I have access to all the as sixes in here so I can maximize the ASIC capabilities.

Even I have 1, 2, 3, 4 line cards And you're building your line cards in such a fashion that if it does, if it's not a 48 port, you know, if you, if you only have four ports, it'll be one per asic, They will still using it. Yeah. Yeah.

Low balancing. So if you look at some of the combo line cut, right? It is not exactly one-four of the ports is one fourth of the bandwidth will be connected to the first asics, right?

Because some of them we have a 400 gig, a hundred gig up link. And then SRP is a down links, right? If you remember on the, on the, uh, the combo line cuts on the 96 0 6, even the same thing in here, the initial line cuts.

So they will have not exactly the number of ports, but the amount of bandwidth. One fourth of bandwidth will be connected first. ASIC one fourth will connect to the second as it's kind of make it much easier for the customer when you're deploying.

I don't have to worry. Yeah, I, I, it definitely goes to some interesting, you know, design patterns in terms of remember where you're mapping ports to. Yes.

A a lot of people do have that question, so that gave them a good idea when I'm, I'm doing, um, connecting the device to it, what port should I use, you know, to maximize the, the capability for on the uplink with what was the links. Um, so those are the line cut on the existing 96 or six line cut. They can be used on this 96 10 chassis.

Okay? There is a carrier module that is required because the form factor of this is little smaller, okay? Um, I, we talk about the um, the power, all the power supply that is on the 96 0 6 is the same power supply uses in here.

The addition in here is we have added an N plus N redundancy. We usually have combined a n plus one, right? N plus.

And if you have two power resource, a power source, I can put one into a outlet power. The other one could be my power generators. So it gave you additional high ties in there.

The fan tray I talk about it is naturally front to back. It's n plus one redundant, okay? Stock wise virtual using the same technology as the stock wise also is a standard base.

Okay? The additional benefit from here just right away is you can dynamically add the start wise virtual link as well as the dynam, uh, sorry, the SVO link as well as the dual active detection links, right? Previously, if you change your number, if you add the links in there, you have to reload the system.

Now you can dynamically add and remove them. Application hosting. This is a huge improvement from the previous one, right?

Start, start out with, well of course faster CPU. We also have more CPU you can dedicate for this memory, but this, the next two component is dramatically different. Okay?

The storage, the SSD now is fuel, um, fuel replaceable unit. Sorry, you push a little button in the front, it pops out. This SSD just like in the UCS, uh, form factor SSD, okay?

Um, the other one is previously if you wanna do an app posting, you have to use a loop back cable to get the datas from the, uh, system data into the CPU so that they can process it. Now we have two internal connections, right? I'm only advertising two internal and we're saying up to 10 gig.

Actually the capability is even higher, but we just see we are looking at the applications, right? With the amount of CPU core that you dedicate to it, you know, can you actually process this much of a power, uh, uh, much of data. Okay.

Um, three more slides. The first one is there was a question on a, uh, unifying onboarding, right? Uh, hey, uh, the Rocky dashboard versus the, um, the catalyst center, all of this 93 50 as well later on with the, uh, 96 10 can be onboarded to both of them, okay?

Natively. So PV today, right? You have to go into the, if you bought a non dash air model, you bought a catalog of 9,300, you go in there and then issue a command, Meraki registered Meraki, and then you connect it to the dashboard, then you can, uh, claim it on the dashboard.

With this, it will be naturally, you can claim on either one, and I'll go through an animation slide in the next one. And if you wanted the switch in between two, you can do so as well. Okay?

When the switch first, when you get the switch, when it boots up, the switch will automatically create a TOS, uh, a, um, a tunnel to the dashboard, an SSL tunnel to the, um, catalyst center and also a TOS to the dashboard. Okay? That's, assume you have internet connections in this environment.

When you connect any of the fund panel port to it and there's already, you have an IP addresses in there or DXCP, he get an IP address, right? Most of our device, when we put in an environment, you have A-D-X-C-P services in there already. So I get an IP address.

This, the switch, when it puts up it initiate a connection to the dashboard. And internally, if you have a catalyst center, PNP running, it will also initiate the connections in here. Now it's up to the user.

If I am traditionally a catalyst center users, I can go into the catalyst center and say, add this device in my inventory, put it in my network. If you do did that, then that tunnel will be terminated. But if I am a market users, I go into my market dashboard and claim this device into my inventory and assign it to the network.

Now the switch basic, say, Hey, I been told that dashboard is where my command is. Then I terminate this part of the connection. So it is very naturally the single device how you wanted to manage it.

It's up to you. Now, you don't have to do any more manual works in here. Uh, I actually talk about this one.

Um, so this is the, we actually talk about this one. This is the, um, with the, right, so we have the PQC sign it in there, and then subsequently for the other, um, features, we, those will we come in the roadmap? A question, Josh, from Diversified, and on your slide here, you have a, a, uh, roadmap in the top right hand corner.

Yes. Since this is like a foundational piece of hardware, just of how the switch works, and this is part of roadmap is do you have any dates that you could share with us on what that means? If not, that's okay.

Understood. Yeah. I might not have the data for the rest of the, so when I say roadmap, it's actually referring to because there are multiple components will come later today.

This sodi, the hardware itself is there, so that one is there. The rest of the component, once we do the, um, the micro loaders, right where this one comes in, this one will come in first and then the rest of them will probably come in around the same time. But the um, the max sec and the IPSec, that will probably at least, uh, I would say six months later probably.

So those will come in much quicker, but I don't have a definite. Okay, so your roadmap is just again referring to the different components of Yes. How you're gonna report, how We gonna do Modules along the way?

Yeah. Okay. You probably will not get all of this at once, right?

Right. It would be like incremental. Once this one's well defined, we have this one and then when the IP sac, when the max sac comes in, and we'll have that additional components.

Okay. Perfect. Thank you.

Cool. Any other questions for me?