Beyond Signal Bars: Optimizing Wi-Fi User Happiness with Arista’s Digital Experience Monitoring

I'm Robert Ula. I'm a technical marketing engineer, uh, for Arista. And I'm gonna talk about digital experience and monitoring, and I'm gonna ask Central, who's remote to also help out with this.

Uh, DEM, uh, which sometimes we refer to as like a digital twin in the old days, even synthetic client testing. It's all the same. But basically, um, what we're gonna do is we're gonna use the multi, I'm gonna sit down here, see no pointer.

Yeah. So we're gonna use the, um, multi-function radio. And for us, that multi-function radio truly is really multi-function.

It has seven functions, one of which is where you instantiate a virtual client, and we're gonna do end-to-end testing layers one through seven, right? From a virtual client out to the cloud. And, and of course it's gonna test a wired in, you know, wifi side and the wireless side.

It's super important that we test the wireless side. Um, and we say that because, you know, if you're just testing on the wire side, right? Um, you're not getting from the client's perspective, you're not testing your radios, you're not testing all your aps, you're not testing even all the VLANs necessarily, right?

And that we know that if you don't test your VLANs, uh, that's where a lot of the problems, you know, can be hiding a VLANs, not trunked on a port, for instance. So it's super important that you test from the, from a client's perspective, and it's kind of think, you can think of it as a golden client because, you know, we can trust this client. We know that it's, it should be able to connect, right?

It doesn't have any driver problems or anything like that. And so, um, you know, it's some, it's a good, uh, uh, test to run as part of the proactive network assurance. So if you look at our dashboards, and I know we've covered this in past, uh, wireless, uh, field days, we have plenty of dashboards, and we'll take a peek at them where you have your connectivity dashboard, your client journey dashboard, then you move up the stack, the performance dashboard, and then finally to layer seven, the application quality of experience dashboard.

That's great. That's all pa that's passive, it's observational, right? This is the proactive part of it.

So we have the two approaches, right? The passive observational, and then the proactive part of it. What's new is the fact that we can do EPLS, um, that was super important for enterprise customers, right?

They, so we have now the ability to, uh, either use the built in certificate on our access points for that virtual client, or download a custom cert from Agni, our own nac, uh, or third party nacs. And so again, it's gonna be client to cloud, uh, multifunction radio, uh, you know, does a lot of things. It does whips, it does R RM scanning it, it does locationing, it does packet capture, it does, uh, spectrum analysis, et cetera.

But one of the main ones here is, uh, DEM. Let me go here. So now I'm gonna turn it over to Central, who's gonna take us through this new DEM dashboard.

And then after that I'm gonna come back and do a little demo. Cool. So I'm central part of the product management team here at, uh, Arista, primarily responsible for the monitoring, reporting, and troubleshooting aspects.

Today, we project the network, uh, network's view of the end user experience by capturing the key KPI categories across connectivity, performance, application, experience, infrastructure. And we also capture the webs security, uh, postures. The widgets over here helps the end user troubleshoot issues live.

And the frame below automatically baselines all the key KPIs over a period of time and bubbles up the anomalies with the red indicator on a click. You'll get to see the clients, uh, impacted and troubleshoot for the, and if you are interested in a first, uh, interested to troubleshoot a particular timeframe, you could use a time range selector to understand how the various key KPI categories were performing. And what's unique about us is our capability to measure application latency on the wire and also on the wireless, uh, medium from the access point.

And all this data is filtered up. Uh, you could filter all this, uh, key KPI categories, A across, uh, the various, uh, SSIDs. And if you have a large network globally distributed, we have a simple way for the end user to look at how the various sites they are, uh, performing across the globe.

And what we have done differently here is provided filters for the various KPI categories. For example, if you're a traditional admin and of, uh, the devices where up, uh, that meets your S-S-L-S-L-A, you have an infrastructure filter. If you're a new age admin and all you care is for the performance of the clients, you have a performance, uh, filter, which will project the performance of all the clients across the various sites.

And if there are clients having poor experience, uh, it would get projected here. And on a click you can navigate to, uh, that, uh, specific client to understand what's going on. We have also facilitated a health, uh, dashboard where the end user can configure their own, uh, KPIs of interest and the threshold levels and the deviations will be projected right here.

Again, all this, uh, uh, beautiful visualization projects. The network's view of the client experience, and this is available at a regional level at, uh, uh, site level. And, uh, the similar fine grain visibility is also available at a floor level.

And this is unique to us. And today, we are introducing something called digital experience, uh, dashboard, which will project the, uh, we are the only vendor to provide, uh, uh, such visibility at all levels in the hierarchy hierarchy, uh, at an org level, at a site level, and also, uh, at a floor level. What I've showed so far is, uh, the network's view of the client experience.

And what we are introducing, uh, now is digital experience monitoring, which provides the client view of, uh, the network. Uh, if you have, uh, retail stores spread across the globe, and there are three, uh, SSIDs of time importance to you, the car SSID the guest in the past network, you could just conduct, uh, uh, configure three test profiles and have them running all the time. So before you, uh, enter the office at, uh, 8:00 AM you could just glance at this dashboard, and if everything looks good, uh, you're good to go.

We are using the fourth radio on our access points to conduct these, uh, tests. These are not overlay sensors. You don't have to, uh, run cables from, uh, uh, additional cables to the switch and, uh, end up, uh, buying more switches.

Uh, and, and it offers a true integrated, uh, experience. In this case, what we are seeing is, uh, the pause application has, uh, failures on a click. You'll get to see the test results.

We know that the connectivity test were, uh, have passed the network latencies. Were, were all good performance, where performance was also good, but we had issues with the application experience on a click, you get validation on the same van latency, land latency, uh, they're all fine. And when you click the application test, you notice that Google Drive works fine, internet works fine, but what has failed is the boss internal, uh, application.

And, uh, to be frank, like we have had this capability all the time for the last 10 years. Now we are coming up with better visualized agents to project the network's, uh, view of the client experience and the client's view of the network, uh, experience. And indeed, we are the first vendor to have provided this, uh, capability.

And I believe we are also the one new vendor today to, uh, project this. Uh, we have also integrated this, uh, uh, digital experience capability, frankly, in the client context and also in the access point context. I'll quickly hand over to, uh, Robert to demonstrate those, uh, capabilities.

Uh, we have a lot more, uh, interesting capabilities in our, uh, portfolio, uh, but, uh, we have limited time. So I'm gonna hand it over to Robert. Yeah.

So, uh, Carlos calls into the help desk and about a problem. Now, I could search for Carlos here and, and find him that way. Um, find his client that way and go to, that's fine.

Um, or, you know, some, some folks like me, when I used to be on the customer side, I used to love to go into dashboards. This dashboard's way more sophisticated than the ones I used to work with. Uh, but, uh, you know, for those of you that remember, um, like, uh, HB OpenView, uh, we have that kind of global, uh, graphical, uh, uh, geographical, uh, dashboard as well.

Uh, that would be kind of like how I start the day and look for anything red, right? And so here you can go to the overview dashboard. And so Carlos called in about, you know, connectivity issue or, uh, some type of a problem, not able to connect to the network.

So this is where, where this is all the, that passive or observational part of the solution network assurance solution. So you have the connectivity, right? Clients as they're going through the journey of getting connected, getting, you know, talking to network services.

Then you've got the performance, you have the application moving up the stack, right? Um, so here it's a connectivity issue. So I can hover over here and I can see we have one issue.

I can this as being an overview dashboard, right? I can go down into the individual dashboards and then see that client journey dashboard. And I see we have a, um, you know, one client that's having a problem, and it is Carlos.

And I can then drill down on Carlos and I get you this really cool, thank you. Centrals, uh, central. Um, this is something that came after Central joined the company.

Uh, this, you know, client is like everything about this particular client, uh, their overview, their connectivity, their, you know, the baselines, uh, KPIs, uh, their performance and application performance, even floor map where you can say, okay, if a client's having a problem, you know, it, it, you know, show me where on the floor plan that problem is happening, right? So you're, you're able to correlate location with a particular issue for a particular client and that, that's pretty cool. So, um, we can see that Carlos is having a problem and then we can do a run a client connectivity test.

Here. I'm going to, I can do that right from here. This is another cool feature that Central helped, uh, make reality.

And we can go here and run that client connectivity test really bad name for a feature horrible name because it client connectivity and by, sorry, we've been, we gotta change it, right? 'cause client connectivity is like such an undersell. It's not just testing client connectivity, right?

Again, it's going from the virtual client out to the cloud and it's testing everything in between client to cloud layers one through seven. But here, um, we can run the test, uh, and we could select the ap. Now you can schedule these tests of course, too, but this is where we're gonna do, you know, a real time or, you know, ad hoc test for a particular client.

While that's running, it won't take too, too long. I'm gonna look at a previous, I'll do a couple things. We'll look at a test that had run.

It's, um, actually sent to, already showed that where you saw the applications and everything. I'm gonna look at the test profile to show you for this particular test that we're running here, take a quick look. So it is ETLS, um, WPA three enterprise.

And notice how we have this certificate tag here. That's just a, a way of labeling the different certs that we're uploading. Uh, you have the default, which comes built into the access point.

Uh, we're not using that. We're using a custom cert that we had previously downloaded from, from Agni for this particular test. Just wanna point that out.

If you wanted to make this test more extensive, you know, you can select all your applications. You can even test, um, you know, uh, captive portal if you wanted to. Uh, yeah, custom.

Custom, yeah. Custom you so you can put whatever, you know, what, are there other applications that you want to test? And, you know, we're going out and doing page pulls for those, et cetera, VoIP, right?

And you get a moss back. Uh, and then also throughput. Let's take a look at the results here so we can look down and we can see, yeah, it's A-D-H-C-P failure.

It's failed. Um, and so we, it is just a way of confirming that it's, you know, it's, uh, there's something wrong with the network. It's not, it's not just Carlos's client, but this is a way of using a golden client to verify that yeah, there's a problem.

And, um, and we can look back here and if we, you know why though? Let's go back to the D Go back here, Leave here, and I'm gonna go back to that client. And this is the key we go back to that we can see here that, ah, yeah, okay.

The problem is, you know, we can look up here and see, okay, that's the root of the problem. In this case, it was a, a missing vlan. And, and so that the fact that there's no VAN there, that, that's, that's causing the problem.

Um, sorry, Pass a question on that because in the information on the screen, it says it was an EPO handshake earlier. Yeah. Which, Which is before you do DAP.

That's Right. Yeah. I'm not sure why we're seeing that error, that this is a perfect example of a live demo not going as well as it went the other a hundred times.

Robert, I, Go ahead. Sorry. Can, I can share too, just in case if you need, uh, some article, Sorry, I, I have the, uh, issue simulated.

So just in case if you want me to share the screen, I could do, I know we are, we are out of time. But yes, if you go into the, Yeah, I think the problem is you're showing three different things, right? You, you see the epo, if you, the bar at the top says missing vlan, then the client says D.

Yeah, Yeah. Which one? Yeah, yeah, yeah.

Oh, well it could have been all three. Okay. So yeah, so, right.

Yeah. But This is the way you, This is proof that all the other times we ran in preparation that it was the right, you know, so yeah. I'm not Robert Gremlin demo gremlin.

I have a question back a little bit. You mentioned you were looking at Carlos's failure. Did you need to choose which AP does the synthetic test or does it pick the closest one?

That wasn't what he, If you had Yeah, I, I'd like to, to be the one that Carlos's client associated to, not the, not have the one he associated to be the tester. Yeah. So how do you pick ap, the nearest neighbor or something?

The ap? Yeah, so you have options usually, right? In our setup, there's only one option because we only had one, one, we have two aps, so you don't get to select from a list.

But, um, when we select, right, uh, Can I answer that, uh, ri here? Yeah. Yeah.

So, uh, uh, Keith, this is, uh, ri I here. Um, so the way it works is, um, the, when you run the client connectivity test from the client's context, in this case car laws, um, so the way, um, uh, that works, uh, Keith is um, when you, when you run this test from Carlos's perspective, uh, the client, the AP that the client is connected to, which is AP one in this case, that becomes the AP on the test. So, um, what that would mean is, uh, when you pick another ap, say AP two or a AP three, AP P three, AP two, or AP three acts as the multifunction radio to run the test against AP one to which Carlos was connected to, right?

So that's how, that's how it works. So yes, you are testing the AP ap, uh, to which this client was very specifically connected. So you, so you're not getting any RF details about how Carlos's clients saw the rf Because you're ing this, not, not with the, with the client connectivity test, you're getting the VIR virtual client, the, you know, that it's perspective of, you know, RSSI and that that's included in the, that's point of view, not, not from Carlos's point of view.

Um, not for, no, you can get that information, but not from the, not from this test. I mean, that's available, right? Uh, but you, for this part, for the test, you can see you have the access point under your test.

That would be the access point that Carlos was trying to connect to. And then you act, have your access point as a client, acting as a client, and you can get, you know, the RSSI there. Um, but that's between those, the virtual client and the AP under tests.

Yeah, I think you've gotta be careful when doing the synthetic tax thing, when troubleshooting an actual well client, because the real client might be, if it was pre shared key, they're typed in the wrong cliche key, so therefore they're failing the four way handshake, key ally tax it with the right cliche key. It's gonna go, oh yeah, everything's okay. So Well, you're not necessarily looking at like for like, by using a synthetic client to tax a real client's problem, um, When, yeah.

Yeah. I think the issue is, and you mentioned it, is the naming that when I'm looking at the client view and I'm doing a client connectivity test, that's not what I'm doing. I'm doing a, a synthetic test from another that's another device that That's Right.

It's think of is a, but I, Yeah, yeah. But I like it as a quick ad hoc test on a, does DHP work for any client that's Yeah. ENS work for any client?

Can I get out to the internet for any client? Um, so you can validate some of that infrastructure before you have to start looking at the specific client that that's the problem. Yeah.

It's like, it's like, so The most important part is, uh, sorry. Um, Robert, uh, I, I, I think two things. One is, um, uh, Peter, to answer your question, yes, the, if the PSK is wrong, of course, right?

Um, but we do have in the dashboard, if the PSK is wrong, um, you know, we can clearly call it out. I mean, based on, I mean we do some, um, analysis of the, you know, pol handshake failures and call out that it disappears K failures. So right of the dashboard, you would be able to see it.

Now the second part is I think just wanna iterate that some of these issues that you see, right? Especially like, you know, say the DCP server, um, is not available, the triple A server is not reachable for those kind of tests. It's really important to run it, you know, from the client's perspective, right?

Because if you run it on the wired side, it's usually on the management wheel lan and the management wheel LAN never has access to the resources or the network resources on the client side, whether it's the application or all the network services. And that's, and, and that's one of the reasons why we've been, you know, investing a lot in building out this capability on the multifunction radio to access point.