During his presentation at Mobility Field Day 9, Rajiv Iyer, a Sr Director in the Product Management Team at RUCKUS, introduced everyone to Cloud RRM. Jim and John had some questions and invited Rajiv to the show to dive a little deeper. What is Cloud RRM, how does it work, and why people should consider embracing RRM in the cloud after spending so much time avoiding RRM.
To see Rajiv speak at MFD9 : Ruckus Purpose Driven Networks – New Solutions and Innovations
Intro music by Alex Grohl, available here:
https://www.youtube.com/watch?v=ZsRWpx8VJ_E
and
https://pixabay.com/users/alexgrohl-25289918/
During his presentation at Mobility Field Day 9, Rajiv Iyer, a Sr Director in the Product Management Team at RUCKUS, introduced everyone to Cloud RRM. Jim and John had some questions and invited Rajiv to the show to dive a little deeper. What is Cloud RRM, how does it work, and why people should consider embracing RRM in the cloud after spending so much time avoiding RRM.
To see Rajiv speak at MFD9 : Ruckus Purpose Driven Networks – New Solutions and Innovations
Intro music by Alex Grohl, available here:
https://www.youtube.com/watch?v=ZsRWpx8VJ_E
and
https://pixabay.com/users/alexgrohl-25289918/
Episode 72 - Cloud RRM
John Deegan: Good morning, Mr. Palmer. How are you, sir?
Jim Palmer: Good, John. How are you doing?
John Deegan: Oh, I can't complain. It's a Friday. We're a little a day late on the recording, but it's fine. It's a holiday weekend. It's Memorial Day weekend for anybody that wants to be completely outta sync when we get this up online.
Jim Palmer: Why do you do this Every time you like, you're like,
John Deegan: it's it's a shtick, man.
Jim Palmer: Killing me, dude. Killing me.
John Deegan: I try.
Jim Palmer: It's, remember this is, this is Thursday, June 1st, not when we're actually recording. We have to time. Yeah. I'll get you there one day.
John Deegan: All right, fine. I'll, right.
Jim Palmer: Okay. Okay.
John Deegan: So, so anyways, the goofiness aside, how are you doing?
Jim Palmer: Good, how are you?
John Deegan: I can't complain. It's, it's it's another day I'm above ground.
And we're recording, which is always fun.
Jim Palmer: Yep.
John Deegan: And today, unlike the last episode, we've got another guest.
Jim Palmer: Yeah. So, All right, so now I'm gonna go back and break my rule that I just imposed on you about 10 seconds ago. So last week,
John Deegan: Rule for thee, but not for me, huh?
Jim Palmer: Last week was Mobility Field Day Nine, and I presented, and there was somebody else who presented IV and he was talking about a new feature that we have.
And so I wanted to get him on the podcast. Now that we talked about it at Mobility Field day nine, you can go watch that video link. We'll put that in the show notes or John will. But we wanted to bring 'em on here and talk about it on the podcast. That way you don't have to actually stare at a screen.
So with that, Rajiv, thank you for joining us. Why don't you give us a little quick introduction about yourself and what is it you do?
Rajiv Iyer: Hey, thank you John. Thank you Jim. Thank you for inviting me here. It's always a pleasure to talk to you. It was very good to see you actually last, yeah, last week and I'll just introduce myself.
I'm Rajiv Iyer. I'm a senior director in the product management team here at RUCKUS, where we do cool radio stuff. And Cloud RRM is another cool stuff and happy to share it with you.
Jim Palmer: Nice. So Cloud RRM, now John and I are, what would you call us, John, we're
John Deegan: words.
Jim Palmer: Well, okay. There's lots of things you'd call us.
John Deegan: Most of it isn't appropriate for the podcast. Okay.
Jim Palmer: So
John Deegan: Wireless geeks, wireless engineers, and.
Jim Palmer: I've had my own experiences with RRM. John, have you had experiences with RRM?
John Deegan: Oh, sure. I've got the scars.
Jim Palmer: So Rajiv I, why don't you walk us through instead of John and I and our jaded, you know, scarred experiences what is RRM and then what is, and then the Cloud RRM, I mean, Can you walk us through that?
Rajiv Iyer: Yeah, yeah, definitely. Certainly. Like the word jaded. Okay. So yeah, we in the industry have seen various versions of radio resource management.
That is RRM. We know that now this version, we call it Cloud RRM because it runs in cloud, obviously. How are this technologies fundamentally different in some sense? Right now we have the compute in the cloud, and we are using some graph AI algorithms to reduce the noise, reduce the Co Channel Interference down to zero if possible.
That is the fundamental goal. We know that. Noise is something that is going to impact your radio networks. Your users are not, will not be able to connect if you have a very noisy network. So the whole goal of Cloud RRM is to drive it down as much closer to zero as possible. For this to happen, what you really have to do is pick channels and channel widths and jointly optimize that. And that's what we are trying to do in the Cloud RRM. That's the high level.
John Deegan: So my turn what makes this implementation of RRM different than others? Because, I mean, like Jim and I said, we're, we've jaded, we've got the scars. We've, we've done this, this is a RUCKUS podcast, but we know every vendor's got their own way of doing RRM.
It's not, you know, it's not a RUCKUS trademarked, you know, thing. So it's been around for a while. So what makes this one different?
Rajiv Iyer: Yeah that's a good question actually. So, It, the times are changing. John, we are going to get a whole lot of spectrum coming our way. We'll get 1200 megahertz, as you know, spectrum in the US and with that comes a lot more complexity and I think Jim covered it really well in the last Mobility Field Day, all the AFC requirements are coming in. And lot more channel like up to 59 channels available new channels in on top of what we had. So the way we have thought about Cloud RRM is to bring help to the Wi-Fi admin already. We all we have 2.4 and 5 Gig and we are going to add the six gigahertz spectrum. We already have some of it in 6E. What we are trying to do here is give them a optimal channel plan, but we know and we have learned that there needs to be guardrails for this type. So even though the AI will come and suggest a channel plan for you, we have built in the guardrails.
And those guardrails are something that I can talk more about.
Jim Palmer: Yeah, so that's, I mean, one of the reasons why John and I are jaded When it comes to RRM is, you know, we've been burned by it. And so one of my favorite stories to tell is I came back into work on a, on a Monday morning one time, and I sat down at my desk and I, and I fire up the, you know, I, I log into the controller and I look and every single one of my five gigahertz radios is set at channel 36 with maximum transmit power.
And I think I had the static channel bandwidth at like 40 megahertz wide. And it was like, it was like, ah, it's like it's been, and I go back and I find out it's been running like this for like, 36 hours, and this is at a facility that runs 24/7. And I was like, oh no. And so it's that point where you just say, I can't do RRM because I've been burned by it.
So I'm really interested. You said that you could talk more about these, some of these guardrails. I'm interested about those guardrails.
Rajiv Iyer: Oh, certainly. I mean, I think the experiences that you, you've added, those are not uncommon. I mean, we have lived through all of those things, right? We've learned from those experiences as well.
So that's the reason we built in the guardrails in the Salgar. So the first thing we thought about was, let's not shove anything down the Wi-Fi administrator. Show them exactly what we are going to change. So there is a prediction, predictive power in the cell goal. So even before the Wi-Fi admin gets to apply that change, they see the change.
This is how the channel plan is going to from its current state to its future state, this is what will happen. And once you have that, you can see visually how the interference is going to go down. So you don't have to accept it, but if the admin wants it, they can click a button, accept it, so you are not sort of forced to make that change.
Unlike the previous Cloud RRM environments, you turn it on, it does its thing in here. The Cloud RRM is always running and the admin has to look at it. And they can decide, oh, this looks like a safe change to me. I'll apply that change. That's the predictive power. The second part is the scheduling power. So you've got prediction, you've got scheduling.
Scheduling basically is just common sense, right? We, we all know that you don't want to change channels during your busy hours, so you, you pick a time. And for dif different zones and, and RUCKUS controllers, you have these zones and they can be very large and they can span different time zones also. So each zone can have a different time, which is a non peak hour, and you can apply that change.
So we got prediction, we got scheduling. Then you also have monitoring. It doesn't mean that we are going to. Let it, that it's not done at that point. What we are saying is we set out predicting that there will be an improvement in the network. We also give you by how much, how much will the interference go down, and then we'll monitor that change for seven days.
We monitor the change and if the changes are not going towards what we said it would, we revert that change. So there is that rollback capability that's also built in. So there's prediction where we tell you exactly how much it'll improve. Then there is scheduling, then there is monitoring, and then there's rollback.
So all these things we're not present in the first generation cloud RRM, any RRM algorithm so far. So that's what makes it different. Does that help you?
Jim Palmer: Yeah. I like the, I like the I, I like the idea of, of, you know, and I think we talked about this at Mobility Field Day. I think, you know, I like the idea of having that, you know, while I'm harnessing and leveraging that, the power of the compute in the cloud, I'm still not just sort of throwing up my hands and going, okay, whatever happens, happens, and then I'm left to pick up the pieces.
So I, I like the idea that, That there is that scheduling, and so on behalf of Wi-Fi engineers everywhere, thank you for, for giving us that ability.
John Deegan: So with, with that said, you know, we've got. You know, improved you know, under the hood and all that, but, so what kind of environments are we seeing improvements in operational improvements?
Because it sounds like, I mean, this isn't hypothetical anymore, right? This is something that we've been using in production environments, right?
Rajiv Iyer: Mm-hmm. Yep.
John Deegan: So, so what, what kind of improvements are we seeing?
Rajiv Iyer: So we have deployed it in production. It's been deployed since December, last December. So we have some real data now.
We don't have to estimate anything. We have real data. We have data and manufacturing. We have data from multi-family home environments. We have data from education environments, some hospitality data, all of that. And what we are seeing is a drastic reduction in Co Channel Interface. And majority of these cases we've seen it go down to almost zero.
So Co Channel Interference is when APs are getting in each other's ways, and basically client has to contend, along with the APs to get access to the medium. But in this case, if you pick up channel plan that is keep giving different channels for different access points, the Co Channel Interference goes down to zero in many cases.
That in turn gives us a better client throughput, higher client throughput, lower their time utilization and higher AP capacity. We have seen all of that.
John Deegan: Nice.
Jim Palmer: I'm actually kind of impressed. John and I were just messaging back and forth. I didn't realize that we had tested it in that many different environments.
You know, I mean, because there all the, all the different environments and verticals that you named, you know, are all. Very unique in, in how the RF operates and how you design them. And so I'm actually, I was like, wow, that's a lot more than I realized. So it's really cool that we've been able to test it in all those different verticals and you know, and realize that it's not just for one specific type of environment.
Like, you know, oh, this only really works in warehouse, or it only works in manufacturing. So that's really cool that you guys have been able to test it that much.
Rajiv Iyer: Yeah, that, that is sort of with cloud, you're able to do that more easily. In the prior generation Cloud RRM algorithms, it was not running in the cloud.
So that was a little harder to achieve that. So right now we can do that very quickly. Oh, actually I missed out high density stadium environments. Arenas. So we got actually very good, very good results there also. So,
Jim Palmer: Wow.
John Deegan: So I, I gotta ask, and this isn't on list, so I'm cheating Jim. So with like, with a, something like a stadium deployment
Rajiv Iyer: mm-hmm.
John Deegan: You mentioned the importance of scheduling RRM, cuz let's be honest, anybody that's ever used RRM if you don't have it scheduled, aside from a DFS hit, it's, it's the worst to have a change in the middle of the day.
Rajiv Iyer: Mm-hmm.
John Deegan: Because it is a disruption. So how does that work with, because, because stadiums are really probably, of all the environments you mentioned, they have to be the hardest to tune, I think, because, you know, if I go in, in the middle of the day, except for very particular scenarios where maybe they're hosting a convention or something like that, the arena's empty.
Rajiv Iyer: Yep.
John Deegan: And the RF in the bowl when it's empty is horrible because the bowls are designed to absorb when the people are there. So how does that work? Like when do you, when do you get to do the channel changes and, and like, right. What's the, what's that like?
Rajiv Iyer: Right, so one of the things that we noticed is when you are learning, algorithm learning the environment, the algorithm is looking at sort of the worst case in some sense because there are no humans in the stadium and you have ha no attenuation between these APs.
So you understand what is the worst case. And if you know that you are able to design a channel plan that says, okay, all these APs are seeing each other all the time, let me find a channel plan that works for that. And when you have a lot of people in the stadium, there is a natural attenuation that you get by just people being there.
So that channel plan that you've already set actually makes. Makes it go better. So some of the high density use cases, we are able to do a lot better now because we actually planned for it already. And channel changes of course, happen in non peak hours, not during game time.
John Deegan: Okay, nice. But that's still, I mean that's, that's a pretty big deal cuz I mean, like I said, those are, those are some really dynamic environments.
So I think I. Any way to throw in that little, you know, bit of AI and whatnot. I think that's that really helps.
Jim Palmer: You know, that's something that I, you know, cuz Rajiv and I were actually talking about this bef right before he presented in Mobility Field Day. And I think that's one of the things that I think, you know, I really liked about it the most was this idea of being able to, you know, sort of.
Take all this information and then predict it and then model it. And if you go back and if you watch his presentation, he actually shows that model of what that sort of looks like. Which, you know, is, is really kind of fascinating and interesting. But being able to predict and go, I'm gonna run 80,000 different scenarios in the cloud, you know, using the.
The compute power of the cloud then come up with and we go, Hey, we can test it all. And so instead of having to, you know, How we do it in the past, which is you'd make a change and then you go, Hey, did this change work? And we watch it in real time. We go, oh no, that didn't work. So now when you make these other changes, so in the past it's always been a bake time, but I love the idea that, you know, doing it in the, in the cloud and having all that compute, you get this predictive and go, Hey, this is of all the iterations, this is I think where we, where we can go.
And so
Rajiv Iyer: That, that's actually a good point because you have the compute, you have a model of your network in, in the cloud. So we are able to run all those simulations, if you will, in the cloud, rather than having to read by hand in a live deployment. So, and that's a good point.
Jim Palmer: So RUCKUS has ChannelFly and I actually did a presentation.
Rajiv Iyer: Mm-hmm.
Jim Palmer: Now this was at WLPC where for a brief second in time, I actually poked a little fun at ChannelFly. And so the elephant in the room now, I think the question is, you know, with Cloud RRM, does that mean that we're retiring or discontinuing ChannelFly or is ChannelFly still going to be a thing?
Rajiv Iyer: Yeah, ChannelFly is going to be there. We are going to support it. And this is something that is our base offering, right? We need ChannelFly because we have environments where APs ChannelFly runs on the APs. It's a local algorithm and it optimizes the RF neighborhood. It optimizes the L link throughput AP to AP, and it has its own merits.
It is simpler in some sense, and it'll just run and it has a settling time when once. The algorithm settles down, you get a channel plan at the end of it, you get the channel bits and all of that is done locally. Now on the cloud. What we are thinking is the design of AI driven Cloud RRM is to tackle the newer challenges that are coming our way.
We are going to get more spectrum. We are going to get ready for all those complexities that are coming our way, and we wanted to move beyond ChannelFly. Along these four, five axis, the predictive power, the monitoring, the scheduling, the rollback, all of that is new. That doesn't come back to ChannelFly.
So ch we are evolving and it's probably the biggest step we are taking since BeamFlex. In radio management, I would say ChannelFly stays. Okay.
John Deegan: Nice. Okay. And that's good to know. Cause I mean, a lot of people, like Jim said, that's, you know, there's, there's a couple things that we've got out there that, that have been out there as part of RUCKUS for a long time.
BeamFlex, ChannelFly and, and and so on. And so it would be, I think it would be a big shock to the system to, to see that. So it's nice to see that the. You know, that's sticking around. So anybody that's listening to this and isn't doing cloud RRM take a deep breath. But so with that said I don't know that I've got much in the way of questions.
I have one last one that I can toss out there, and Jim, if you've got anything else we can kind of follow up on that obviously. But Rajiv, what is like one thing if I'm the listener and I am a listener what would you want me to take away from this conversation?
Rajiv Iyer: Ah, okay. I think AI driven Cloud RM is the biggest advance that we have done since BeamFlex, and it has four properties that are very, very different, which is predictive power, scheduling, monitoring, and rollback.
All of these four properties of this cloud RRM algorithm really distinguish it from the past attempt that we have made, that the industry has made, and that's what gives it the power. That's why it needs to run the cloud. Do, do, give it a try.
Jim Palmer: All right, so. You, I do have another question cause I actually, I, I actually take notes.
I have my little notepad here and I take notes about what's said. But you, you said something early on that I want to kind of come back to.
Rajiv Iyer: Mm-hmm.
Jim Palmer: Now I thought you said something about a graph AI algorithm or something like that. You, I, I wanna know more. I mean, I think we have about 10 minutes left, so you got plenty of time.
But what exactly, you know, when you say graph AI algorithm
Rajiv Iyer: mm-hmm.
Jim Palmer: Is it, or, okay, first off, is that what you said?
Rajiv Iyer: I said graph AI algorithm. No kidding.
Jim Palmer: Okay. All, so two people like me and John who are again, old fuddy duddies. What exactly does that mean? What is a graph AI algorithm?
Rajiv Iyer: Alright. Alright, let's break it down, right?
The graph networks, like you have a bunch of access points deployed. You can't see it all in one room. I'm, I'm sitting here and there are probably 40 access points in this building. I can see eight of them, but these access points can see other access points. So it is actually like a node and a graph and you scale it up.
You have multiple different buildings, multiple different zones, all of these form. The best way to visualize this is a graph. It's like now these nodes are also divided in different spectrum in 2.4 gigahertz. Some radios are seeing some other radios and five gigahertz, some radios are seeing some others.
If you walk into an environment, you get a sense of saying, huh, that looks like an AP. Too much there, or There is no AP there. That's a dead spot. You as a RF expert have that intuition, but obviously you don't want go through every single facility to figure out where the APs are deployed in high density or maybe their dead spots.
So if you. Give AI this graph that we built and that graph is learned by the data that we get from the controller. It is able to tell you what you are seeing humanly when you walk into a node. It will tell you exactly what the clustering density is in what places and where it is not. So it learns all that.
So that is the graph part of it, and the AI will help you. Sort of transform what you know as a gut feel into something that's very mathematic. So we got that part. So now we can make some good decisions based on where the dead spots are, where are some high density areas, where are some low coverage areas, all of that.
We can find that. Then the part interesting part is to see if there exists a channel plan. That will make these APs not interfere with each other. The Co Channel Interference. Now, if you think about it, that's actually a hard problem because you can see one AP and that AP sees another AP, and you can say, Hey, that's on channel six.
I'm on channel one. Let's just that there is no Co Channel Interference. If you see this AP is on channel one, the other AP is channel one. Let's change my channel to six. But there's an AP that you cannot see that's also on Channel six. So that algorithm, if you want to scale it up, it has to be run in the cloud.
And now you are creating, coming up with a solution, using graph AI, graph AI, which will say, I have now run through those 80,000 different scenarios or more to give you a solution that is. Driving down Co Channel Interference to as minimal as possible. Zero is what we want to drive to. That is what the graph AI technology is doing really,
Jim Palmer: And for those who are wondering, when you says sees the other APs talking about an RF level, which you know is, which is, which is really, is really hard because you know, Well, I was in the office last week and I happened to look up and oh, by the way, there's more than 40 APs in that building.
That's a, that's a whole other topic, but you know, and for a trained person it's e you know, it's not easy, but it's, you're a little bit better at going, okay, this is, you know, I'm pretty sure that this type of interior wall is. You know, just 3 dB so it's like, yeah, I'm pretty sure that they, this AP in this room and this AP in this room can probably, you know, from an RF perspective, see each other.
But, you know, a lot of times we don't, you know, you might think that it's a 3 dB wall, and so that's where that seeing part comes in. It's because it, it's taking all of this stuff, you know, that the APs learn about its environment, and that's how, that's, that's where the seeing part comes in. For those people that don't.
Aren't quite sure how that works. It's actually a, a sounding thing that happens, you know, in the background of the APs as it learns its environment, which actually as I learned after Mobility Field Day, also ties into AFC. So this whole, you know, learning and understanding where the AP lives at an RF level within the environment that's deployed is actually quite critical for several different features.
So that's, but, and. Is So the graph AI algorithm last follow up to this, that's the, the, the graphs that you were showing in your presentation?
Rajiv Iyer: Mm-hmm. Yeah.
Jim Palmer: Okay. All right. So yeah, when you go watch that video that's the, the little little charts that he was that, that he was showing off. So, all right, so that makes more sense Now.
I'm good. Well, I don't know about good, but I.
John Deegan: You're good for Jim. I mean, I really don't have, I think too much for this one. But it sounds like, well, I'll add this one last thing and, and sort of a comment and to tease it up for a, a future return visit for Rajiv as we like to have our guest back.
Let's be honest, we have a, a small family of RUCKUS employees, so we've, you know, we've gotta keep everybody in the loop. But, so with it being in the cloud, with it being Cloud RRM, how, like, is it on any sort of an updated schedule? Like is this gonna be tweaked and tuned and, and fine tuned on a, on a fairly regular basis?
Or what's your expectation of like 6, 12, 18 months? What's the, what can we expect from
Rajiv Iyer: it constantly working on it, John. I mean, as we, this is, this is cloud delivery, right? So we, we try not to shove in a lot of changes, big changes and things like that, but it's constantly being worked on.
John Deegan: Nice.
Rajiv Iyer: You will see a lot of good things come out with AI driven Cloud RRM and in general, we think that's the way to go.
The complexity of this network is going to only increase the spectrum access rules are changing. The rules of driving on the road are actually changed, and most of those things are making, making it harder. I mean Jim talked about the MLO. Dynamic link selection. We do not want to transfer the burden of all that to a human Wi-Fi admin, right?
So we are going to get some algorithmic help and AI driven Cloud RRM is our solution.
John Deegan: That's good stuff. Yeah. That's all I have. Mr. Palmer, I don't dunno if you have any other,
Rajiv Iyer: Why don't
John Deegan: Questions to us
Rajiv Iyer: Give you an extra piece of information? How about that?
John Deegan: Oh, there we go.
Jim Palmer: I love that.
John Deegan: I, I like it. He's pulling out the Steve Jobs. One last thing I like, it's
Rajiv Iyer: More than one last thing. I'm, I'm just saying that I did say that it runs in the cloud. In the cloud, in RUCKUS Analytics, and it does require a license, right. Now, many people come and ask us what happens if the license expires? Will all my channels reset to whatever it was before?
No, we will not do that. So we will take a very responsible decision. Whatever cloud RRM ran last time, whatever is the result best, whatever we know was the best that will stay. And then the channel plan will be at that point handed over to the admin and we will not reset it back to basics. So please don't worry about that.
Jim Palmer: Yeah. Yeah. I'm glad you mentioned that. Cause I'm pretty certain somebody listening to this is gonna add is thinking that as we're talking. I'm like, wait a second. And I wouldn't think to ask that. So thank you.
John Deegan: Nice. Good to know. Well, I, I. Unless Rajiv's got another you know, surprise out of his, out of his hat. I don't think I have anything else. Mr. Palmer, do you have anything else to, to cover, ask, bring up, mention you?
Jim Palmer: Well, you know, you, you were talking about, and this is still such, I mean, to me, I've heard about this for a while, but it's still sort of such a new thing that I would like to ask Rajiv if he, you know, if in six months or whatever as we've had time to sort of think about this and learn more about it, if you'd be willing to come back.
And discuss more because I mean, like I said, John and I have been chatting and we have some notes and it's like, well, what? And it's, and so maybe we can come bring you back another, another time and and we can do a little bit more of a, maybe a focused on just one or two key points and, and do a little bit deeper dive.
If, if you're okay with that.
Rajiv Iyer: Well, what date is it? Five. 26. Right. So how, how about we talk, talk back in a couple of months maybe? Or let's say a quarter. Yeah. In a quarter. Let's,
Jim Palmer: so, well we gotta figure that out cause we're, we're busy people. We're in demand here. But no. But if you're willing to come back, we'll get you, we'll get you back in and we'll do that way we can do.
You know, a little bit deeper dive and have some better questions and better notes. But yeah, if you're, if you're game, then we're game. So thank you for that.
John Deegan: Well, cool. All right, last call. Because every time I said it, we've got, which is good. I like it.
Jim Palmer: We've been last call for like five minutes now,
John Deegan: but no.
If, if there's nothing else, I don't have anything we can call it a wrap for the day. Rajiv, I wanna thank you for taking time outta your busy schedule. I know you know, being in your position, it's you've got a lot of demands on your time, so I appreciate it and yeah, definitely. Thank you very much.
Jim Palmer: Yes, thank you.
Rajiv Iyer: Yay. Thank you for having me there.
John Deegan: All right. Well, if there's nothing else, Jim, Rajiv have a great well, officially, as Jim said, it's June 1st, so have a great rest of your Thursday. Wink, wink. No, no, but enjoy the weekend. Don't celebrate too hard. Relax, enjoy the time off and yeah, we look forward to having you back on in a few months.
Rajiv. Yep. All right. And we are music.
I had to find the stop.