So specifically when we talk about our SDW a solution that we've been working on together, um This topic of DMPO comes up a lot. What is DMPO? So that's another great question for our customer journey. So I'll start off by writing that acronym up here. And maybe you can speak to that DMPO.
DMPO Dynamic Multi Path optimization. When I got started in uh SD wan years ago, uh it was one of the huge, you know, um uh lightning rods that kind of hit my head when I realized that we're in the dawn of a new and a new era if you will. And so to help explain it really quickly, let's start off with a typical branch, right? That has two links and they could be both internet or they could be mpls or both mpls could be any combination of the two.
In fact, just for argument's sake, let's say between my branch and my data center, I have three different links. OK. So the biggest problem we have today in modern routing is whenever I go to set a path for a flow of traffic type right. I am essentially to statically assign that application to write on that link until something breaks and then I want it to move. Unfortunately, today, for most routing, that means when something breaks, it means it has to go completely down. It can't just be behaving badly, it has to go completely down. So the first thing that we want to do with SD one is we're going to front end all of these devices with our SD one service.
And as all the applications are coming in to the service, I'm going to be able to identify what they are, what is the actual application and not like back in the day when we were just, we could just figure out what voice is. We're actually going to identify what all the applications are, you know, is this a Skype Messenger or is this Skype voice video? Right? OK. So now if I know what's going on, if I know what the application is, now, I need to look at the underlay, the underlay of the links and the three things that we really want to try to look for, right? Is how much latency I've got, how much uh throughput I have put TP there, right? Uh And if there's any uh packet loss or Jitter, OK.
These are the four things that will kill any application. So with every single packet that goes across every single link, we are going to put a time stamp and a sequencer on it. Now by having a time stamp and a sequencer that means that as my packets traverse the network and they come to the other side, the other side is going to say, what time did that packet leave the other side? How long did it take to get from point A to point B? And then every 24 hours, we do a bandwidth measurement on each one of these links. So I know what the latency is. I know if whether or not I've got any packet loss and I know if I have any jitter with every single packet I receive on the other side, this means that when I receive these packets and now I'm going to present a new application to this, to this underlay.
I'm doing so with the knowledge of exactly how that link is performing at that exact moment. So I'm doing real time, right? Acknowledgement about what applications are going well, which is, which is just crazy for me. I came from a world where voice was absolutely paramount. Voice calls absolutely could not come down. So what's really awesome about this is that, let's say that this link here is 100 megabit link and it's operating beautifully and all of a sudden boom, it goes completely offline. I needed a solution where I could look at this link and I could say, hey, it's gone offline. I need to move that voice call, right? These voice packets from one link to another. And to be clear, we're not talking about you hang up and redial in and I'm now on this new connection, I'm literally in the middle of my call.
You know, my next word out of my mouth has moved off of this link over into this link. Absolutely. And it was the most brilliant thing when we saw it when we were actually testing in our own labs, how well that worked. And that's when the light bulb really went off for us. So now we've got the solution where I'm taking a look at the underlay every time I want to present a packet to it. I'm doing so with the knowledge of exactly how each one of them are performing. Now, that's great. And you're saying to yourself, well, Greg, that's awesome. If one of your links is running great, even if the other ones are not working. But I live in the real world. I mean, sometimes this one here can be impaired and this one here can be impaired and this one here can be impaired.
What if they're all impaired? What if I'm suffering from latency or packet loss or jitter in any one of these links? What are you going to do about that? We're going to do a couple of things here. Number one is we're going to allow ourselves the ability to do packet duplication for real time applications. Now, real time applications are very sensitive, you get one shot to move the packet from one part to the other. So what we do is we will duplicate real time packets. When we see that there's a problem with the link, the link is behaving badly that gives us that doubles the chances for the packet to get from point A to point B. And we have seen in our own testing up to 12% packet loss, still maintaining a good voice call, which is just crazy in this state. Unheard of. One of the other things that affects real time applications is Jitter.
Jitter is when the packets are coming in out of order, right? Or if the packets are coming with a delay in between each. So what we do then is we just put a little bit of a buffer on each end that allows us to reorganize the packets and put and present them to the application. Now, the beautiful thing about what we're doing here is we're doing all of this within the W A. The actual applications have no idea, right? The user and the, and the, and the actual application have no idea any of this is happening. So as far as they're considered, the entire network is running beautifully. Now, there are going to be some cases when all of these are running horrible or let's say that this actually down here is an LTE, right? It's a, it's a cell backup, right? And it's used for those cases, when your main primaries are down, what we can also do is because we can recognize what every single application is, right as it's going through, we can create a prioritization list.
So we can say at points of contention, elevate these and let these through and these here drop packets for so that we can put only the only the good applications on that link my Skype calls and my ERP system at the top of the list. Yep, Facebook, youtube cat videos at the bottom. They are getting a horrible experience, right? But the great thing about this is that you don't have to continue to redo this statically. All of this has happened automatically based on the policies that are set up by the administrator. Well, yeah, I think that's been a great uh you know, coverage of this whole topic of the NPO and why our solution that we're bringing to market around Sdwan is fundamentally differentiated. I appreciate you taking the time to listen today.