Arista and Wi-Fi 7: What the Real World and Arista Lab Tests are Really Telling Us

So today from, um, and, uh, MRO after that, we are gonna deep dive into learnings from the field. I'm ashwin, I'm part of, uh, wifi, TMET. I mainly work on, uh, wifi solutions and field enable enablement.

So, and today we are gonna deep dive into, uh, learnings from the field. Um, so we know, um, when, uh, for the very first time, uh, when the wifi seven launched, right? Mlo, MLO was one of the key features of, um, uh, the wifi seven.

So we, in fact, in the last, um, uh, mobility field day, we kind of showed, uh, three very interesting demos. Uh, one is, um, we showed the spectrum view of SDR ML R and then we talked about the spectrum view of EML sr. And then we also talked about preamble punctu, right?

So in this session also, we have two very interesting, uh, uh, live demo. Uh, one is going to be, um, uh, spectrum view of E-M-L-S-R in, in terms of interference. And another one is a, a live, uh, MS teams call, right?

So, and then we will walk you through the day two operations on what kind of benefits, uh, when this is going to happen, right? So, without a further ado, let me jump into the MLO modes. There are, we have, we have seen this slide multiple times.

There are two radio solutions, single radio, multiple radio solution, but, uh, what it really matters down to the client is, um, uh, the E-M-L-S-R and then the S-D-R-M-L-M-R, right? These are the two, uh, modes, which is, uh, which is available at the client site. So what is SDR ml?

MR? It is, uh, what is a benefit? It is bringing up to the table.

It is bringing up, uh, it is giving us the aggregation benefit, right? Where that, where we can, like, you know, concurrent, we can do concurrent TX RX over the multiple links. In the case of E-M-L-S-R, uh, it is bringing us the steering, uh, benefits across multiple links.

Uh, what we see here is the, the client is going to have, um, like, you know, the tune the single stream across two, uh, uh, two chains. And the, as in, when that is a packet downstream to the ap, it kind of indicates through, uh, the control frame, M-U-R-T-S, and then the both, uh, multi both the chain get, uh, all the client gets switched into, uh, uh, to receiving the two stream data packet, right? And this is, uh, this will bring us, uh, the steering benefit.

And, uh, what we have seen, uh, uh, over the period of time is, uh, we need to have a proper SSID strategy, right? 4 5G and six gigahertz band comparing to having one on five and six, right? 4 and six, it connects it in, uh, as an S-T-R-M-L-M-R.

4, uh, band will get, uh, highly, uh, have and high interference and, uh, it's a polluted channel overall. 4 and five. That's what we have seen.

And, uh, in last, uh, MFT, uh, we kind of presented, uh, the, the client MLO, uh, mode selection list. We just added a few more clients, uh, especially Google Pixel nine, um, iPhone 16, and, uh, 16 PRO has come. So we also had added the table for chipset and, uh, TR ml, non-ST R ml MR.

What is so interesting on this is when you have SSID on all the three bands as we discussed, it does STR ml, MR, and, uh, if you look into the iPhone chart, uh, at the, what it actually, when you have it on five and six particularly, it is advertising a non S-T-R-M-L-M-R, okay? But it does not advertise EML sr. So ultimately what happens is the data transfer would happen only on one link.

45 and six, it does advertise as the S-D-R-M-L-M-R, but, uh, data traffic would always happen in one link, right? So that is the key, uh, observation we had. And, uh, rest of the observations are actually in line with last, uh, uh, data that we published.

Uh, in the last call, uh, now we will deep dive into the real demo. So in this, uh, demo, what we are gonna do is, uh, we will have a Google Pixel eight, uh, participating in EML SR mode, and then we will strike the interference on five, uh, six gigahertz band. And then we will see how the client dynamically switch over between tool links.

So in a high level, this is what our setup would look like. Uh, we have, uh, our seven 10 P switch connected here, and, uh, two aps connected to the seven 10 P switch. And then we have an interferon wireless interferer, uh, also connected to the same switch.

Um, and then, uh, we, we will be doing, uh, MLO, uh, so what we see here is, uh, we have SSID one only on five and six, and, uh, this client is connected, uh, in both five and six. As you see, the frequency is highlighted as five and six. Now, uh, when I ran a speed test, okay, so, uh, okay, so I think the interference is already selected.

Uh, no, we, we are seeing an energy here at, uh, six, maybe you can turn off the interference and, uh, we can try it again, right? Are you doing doing demo two before demo one? Yeah, the demo Demo two are not shining today.

Okay. So, uh, I think the interference is already there. That's what the, the, the traffic is going here.

Uh, but the same test, uh, I kind of like, you know, um, had a recording of that, uh, to be in a safe side. So, so I'll just walk you through this recording, which is exactly the same because this test, we ran it in the morning. So what we see here is we, I'm just, uh, showing you, uh, there are like, um, the, the APA radios are there in one point in all the radios, and then these are all the list of SSIDs.

I have it. And, uh, in this particular test, I'm gonna strike it because it's the recorded one. We are gonna strike it at, uh, interference on five, and then see it, uh, moving to the six, right?

As you see, uh, we will be first connecting the client. So the client, uh, what we saw is five and six. Uh, if you see on the spectrum, the, the spectrum is completely free, uh, on the five gigahertz and, uh, six gigahertz, right?

It is free. And then I'm starting the speed test. And then this time you see the good throughput here.

And then what we see is the energy picked up in both five and six, right? So now you, you focus on that, uh, that area of 1 53 channel space where there is a, a spike of that. Once we inject the interference here, uh, what we would, uh, see shortly is, uh, the traffic at five will get stopped.

See this, uh, the interference strike, right? So you would see the client smartly moves to the six gigahertz without disconnect, right? So this is one of the key, right?

In the last MFT call, we kind of striked it on the secondary channel, right? And then we showed the benefits of MLO, uh, MLO when it was dynamically like, you know, adjusting the bandwidth and it was sharing. Now we are striking it on the primary channel to see how the client dynamically switch to the other band without any difference.

For the legacy system, what it used, used, what would happen for the legacy system is basically it kind of, uh, like, you know, disconnect or like, you know, move to the different channel. So this is a key benefit. That's one of the reason why we need, uh, to move on to wifi seven, take the benefit of MLO feature and start moving on to this, right?

So this is this part of the first demo. Uh, and then the second demo is, uh, even more challenging. Uh, so what we are trying to do for the second demo is, um, we, we are trying to bring the day two operations live.

Right now, we will have two clients. One is MLO client, and then the non MLO client both joined the same access point, Right? And then the active teams call will be going on.

When the teams call is going live, we kind of bring in the interference we wanted to showcase the team's call will continue to go on for the MLO and then for the non MLO clients too, at the end of the call, we kind of give the summary, right? The call quality that we having the ability to take the call live and then what is a call quality? The call quality on the MLO client will not, uh, because, uh, the client would smartly move to the other band, whereas the call, which is there, uh, for the wifi, um, uh, the six client, which will deteriorate, and then we should have the ability to detect that and send the live feeds to the admin administrator, right?

So even before somebody actually reports a support ticket, we should be able to see that live, right? So that is the goal of, uh, the this, uh, second call. So what I'm doing is I have, uh, uh, three participants here.

One is, uh, uh, me, Alex, uh, I, I will be the presenter of what you see in my name here. And, uh, another client Alex. Uh, I want one of the volunteers, uh, to take this, uh, uh, pixel eight phone, which is an MO phone.

We want you to like, you know, show your, uh, phase and see the real time, the benefit of Helo. Uh, so I will make the call and then the, um, and then the next, uh, um, the person would be Bob, uh, would pretend himself as a Bob, and he will also join the same conference call, right? So you can have, uh, uh, one of you can have this live, so already are live.

So sorry, Ross, gimme a smooth look. So, so what we see is we kind of already had the interference. We were not able to stop the interference, right?

Uh, in the previous demo. So this is the same experiment that we wanted to do live, right? So this also has been taken live.

So we have two user, Alex and Bob. And, uh, what we are trying to show is both of them are connected, uh, you see and, uh, MLO, uh, which is a master MLO link. And then we also have a link view and, uh, the master view.

And, uh, we are bringing in, uh, the teams call. So same, uh, was, uh, acting as a bob here. So, and uh, what we see is, uh, he is already there in the call, and then I am there, but with a different dress in my office trying to explode this.

And, uh, what we see is in the lower side, I am there on the top, and then I'm showing my own face, and then bui there is in the other side. So he is waving his hand, so all good. And, uh, we will, uh, very shortly induce an interference, and then what we would see is boop, in this video will get completely stuck, Right?

Right. So in at this moment, we have already injected the interference, and uh, this is how, uh, we wanted to show the spectrum what you see this in this test, it was already operating in channel five, and then the interference is stuck on the five, right? So what we actually see is the two clients, uh, there is no impact on the MLO client, whereas there is an impact on the non MLO, right?

That's what we wanted to show. Okay? Gotcha.

Okay. So remaining This, this is the E-M-N-S-R, This is on the E-M-L-S-R, right? Yeah.

So if I just fast forward a little bit, and there comes a very interesting, uh, view from our CVQ, right? Um, Right. So what we are trying to, uh, show here is the live feeds, right?

So we have the ability to detect the call. As you see the mlo, uh, the non MLO client is suffering, but the MLO clients are not, right? And then there is a live feed for an administrator where you get, uh, the live, uh, in, um, there is one client which is actually suffering in this test, right?

So you can, um, and this has also been integrated to Slack and other, um, MS teams and, um, other web books. So what we see is a poor application being detected live, and uh, we can, we have also built the explorer view that clearly the calls out, like tells how long the call has been running and things like that, right? Are you getting the, the, the team's quality info from teams or are you No.

Doing, Looking that through? No, we have our own, uh, stuff in, uh, we, we did, we did train a lot of, um, packets and classified our own algorithm for That. Okay.

So it's not a teams integration, this Is not a teams integration. So When you say I'm looking at the live feed, how are you, how, how are you, how do you know, like that's the call that both of those devices. So, So we have our classification engine.

Mm-hmm. So that detects what, uh, so what call it is, if it is Zoom or teams, we have the classification engine built internal, um, uh, and then we kind of classify them, right? Yeah.

So I also mapped this classification of teams call to an access category of voice, right? So that's where when you saw that video though, the Alex, uh, was not the, which is Google Pixel nine, which was working fine. Mm-hmm.

The Bob, which is acting as pixel six, was not disconnected on the call because we are already pushing that into the voice access category. Yeah. So if we have a group of people who are on a call and you've got the one guy with a device that's not, you know, yes.

Uh, wifi seven and MLO enabled, you're able to go into the dashboard and say, yeah, only these specific people are impacted, but we've got newer devices that are still able to be on that call. Yes. And so I can get that information here from that.

Correct. And not only that, when you are actually on the call, when you are suffering, uh, badly, you'll get a live feed saying this, right? So here you see one client Bob faced a wifi issue that led to the poor application quality at that location.

Uh, and then these are all the possible root cause of that, right? So in this case, the root causes are three because like, uh, we are running a interference test, which is not practical, which is loading your network for 95%. Yeah.

Okay. You, you gave this demo previously, which is great that when you kick an interference then MLL does its magic. I was wondering what is the, that's a pretty obvious thing.

I mean, you're just blasting that channel in the real world. A lot of times the downside is a channel's busy. What's the trigger to make MLO go?

So I mean, you, you literally took out a segment of, of the frequency. Yeah, Right? Like under congestion, Like does MLO trigger with just congestion?

I mean, there's other things that would be where MLO would solve the problem. Uh, so it is, uh, if I hear, hear your questions, right? So you wanted to basically like, you know, uh, the trigger to stop the MLO is that to Start, The trigger to start is it's all based on the, the, the CCA of the individual, uh, the access point on that client based on certain threshold.

It, it knows that which channel is free, and then it just kind of like move that traffic, the other other band. I guess the question is, will that CCA trigger with just congestion? Yes.

Or does it have to be It is on the per packet. It is on the per packet basis, and then it just, uh, see whichever link at that moment is free and then it kind of send it to the other other channel. Thank you.

Yeah. So, um, yeah, so that's all I have for the, uh, the two demos. Sorry, we could not, we, the demo gods were not favoring us, but then we kind of, uh, covered it with the, the demo, uh, recorded A one.

Yeah. I have a question about the, uh, you know, the STR versus the EMMR. Um, you said in most companies nowadays they deploy on five and six, right?

4 as well to take advantage of the SDR minimum? So, Uh, yes. So it's purely based on the admin administrator's, uh, thing, right?

4, we would certainly get 10% gain overall K, right? 4, but the actual practical deployment, what we see on the field is going to be on five and six, right? So, so in that way, we been asked to be aware that, okay, in five and six, this is what the E-M-L-S-R can dynamically, uh, switch between links and you should get the benefit of, without disconnecting the client, how I can switch between tool links.

That is a key, uh, aspect Here. Yeah, it's more about the reliability of the connections. Reliability and latency of the, Yeah.

Okay.