Today we are here with Sakid Ahmed from Cambium Networks to discuss QoE, Quality of Experience.
What is your role at Cambium?
I am vice president-general manager for the EPMP business, but today we’re going to talk about the quality of experience and you guys are releasing a product now called QoE, or quality of experience.
What exactly is QoE to you?
This is not a new term, it represents the experience of users at home. Whether it is us, or you guys at the distribution, we are in the business of selling products that enable broadband service. No matter, if it is Wi-Fi, switches, access points, etc very rarely, do we think about the content that has been delivered, and how well is the as well content working. What is the perception of the end-user? If you think about the modern home and all the things that we have running from Google Five to Amazon to Nest thermostats to Ring doorbells, let alone the web conferencing and the applications were running as a service provider. Whether you're a WISP or an MSP it doesn’t matter, how much emphasis are you giving to the end user's perception of the application they are running? It has everything to do with choppiness in your voice audio, Netflix buffering, to high latency on online gaming servers. What can you do besides just going for higher bandwidth? Going for the larger channel sizes, going for fiber, etc. What can you do to change that experience, because that is the essence of QoE? That matters more so today than ever before. We are shifted now quite a bit more to that home user running a home office, to gaming, to Netflix, everything is sitting together on the pipe. Can you improve that experience? What more can you do about it? That’s what Cambium does.
If you think about Cambium traditionally we’ve been talking about the radio side of the house. We are the guys that preach about MIMO (Multi-In Multi-Out), how can we get more data through the pipe? This is another angle. We’re good at getting data, what can we do more with that? That’s kind of the essence of QoE.
When you think about laymen's terms we have talked to a lot of WISPS today and they’re like I got a ton of gamers, we got people working from home now doing video conferencing like we’re doing today, etc. and we are already creating those pipes to the homes, but now it’s basically like creating a better pipe without actually doing more. We’re breaking into it and making a better experience for the end-user without changing the hardware?
That is exactly right. You put one of our appliances in the NOC and you pipe all your traffic through that so it’s a combination of things like a DPI engine (Deep Packet Inspection engine). I can give you a used case for example: let’s say to our DPI engine you’re looking at three days of the history of all the data that’s going through the pipe, your network, and you say "holy cow" on a weekday for some bizarre reason between 8 PM and 11 PM there is a lot of BitTorrent traffic. A few users are using a lot of bandwidth. You don’t know that data today, you just know that somebody’s using a lot of bandwidth. You don’t know what it is, but now you can go and say all right fine there’s a lot of BitTorrent being consumed so I am going to create a shaping rule and squeeze the BitTorrent down to maybe 5 Mbps. All of a sudden you took out one of your biggest offenders that are using data for BitTorrent and by choking it you’re gonna allow the Netflix users and gaming users a better experience. Shaping based on application. You can do this for BitTorrent. If you take an example of Apple TV and the NFL Sunday ticket you can flip the story on that. Sunday night your users are gonna watch NFL, do you want BitTorrent to eat the bandwidth? No. Give a nice solid pipe for Apple TV, you can detect Apple TV, open up that pipe and squeeze everything else. Do it on a time basis. Come Monday morning that rule expires and a regular rule comes in. So that’s just one aspect of QoE where you can do application visibility and shape in order to create a better quality of experience.
In speaking of the QoE what features can you do with this device? Obviously, we talked about rate limiting, but what other features are available in this QoE?
I will list them one by one. First and foremost is what we call TCP Optimization Feature. It’s a proprietary TCP proxy. When a QoE box is sitting in line it is responding on behalf of the TCP server-client with acknowledgments. What we can say confidently is whether you're running a fixed wireless network, or a fiber network with a bunch of WiFi SDN or an MSP doing a senior citizen home you do have interference, Wi-Fi will have problems, and packets will take time to deliver. TCP, which is the dominant protocol these days, by nature will look for problems and then try to correct it by backing off and cutting the throughput and all of that. So our job is to sit in the middle and let the TCP server think that everything is great. We’re making an acknowledgment on behalf and saying "hey, I got the packet, don't worry, keep sending data, don’t throttle back". By doing this TCP proxy you are going to increase the bandwidth of your network and get the application not to start buffering and slow down. That I would call one of the most anchored features of the most differentiator features you don’t have to do anything with. You simply turn on the button and say TCP optimize. So if you are a WISP with a tower and you have a bunch of radios where you get a lot of customer support calls, there is a lot of interference, you are constantly looking for channels get a QoE box, turn on TCP optimization and let it do the magic on its own. You don’t have to do anything.
Then you have your very granular application shaping. Being able to see what’s traversing the network and being able to create rules. If you say I wanna squeeze BitTorrent, I wanna squeeze Netflix, I want to squeeze YouTube, Hulu, you name it. Hundreds and hundreds of DPI signatures. A customer the other day wanted to be in Peacock as a signature and DIRECTV stream, we added that within seven days of the request. A rate limit by itself is an interesting thing. You have traffic shaping based on application but rate-limiting becomes where you have an SLA of 50 Mbps down, five Mbps up. There are companies that do rate limiting, but you can also do rate-limiting at the AP and SM, but there is a very harsh packet-to-drop mechanism. If you get to the 51 Mbps packets are getting dropped again. In our case we have multiple queues inside the rate limit pipe, each queue can be associated with a particular traffic type. Netflix has its own queue, Zoom has its own queue. When you reach the mass capacity of the pipe (i.e. the rate limit) you can start dropping packets on the different queues, as opposed to one only queue. Now you’re balancing the user's experience while dropping packets is not as harsh of a situation.
While we’re talking about rate-limiting we just added another feature that is the ability to detect speed tests. Now, what happens ifs if Jeff has a 50 Mbps downlink plan in his household the kids, or wife, are eating up 30 Mbps of the data. Jeff says I’m gonna check, I’m paying this ISP $99 a month so I am going to hit the speed test to see if I'm getting my 50 Mbps. Now when you hit the speed test the QoE appliance will detect that Jeff has initiated a speed test. It will keep all those applications in your house running with the necessary bandwidth and will give you a bursting. All of a sudden you are going to see the full 50 Mbps as the result of your speed test. You are not gonna call your ISP and complain that you're not getting 50 Mbps. It makes a lot of sense because everybody complains about that speed test. That’s rate limiting. You have your granular traffic shaping, you have your TCP optimization, and you have a ton of stats and full transparency. In this space, you get companies like Preseem, great companies that did a lot of good work for our Wisp industry. iYou have some big boys in the network and you got a few other players, we are one of those companies. Our differentiation is unlike a pure QoE company we are both a QoE company and a Cambium hardware company. If you use cambium we’re going to do some integration, where we can do things like multiuser MIMO gain announcement, understanding AP thresholds, and having QoE do things automatically. The good news is if you’re not Cambium it doesn’t matter, you can put a QoE box and use it anywhere you want. That’s one big advantage.
The last piece, along with visibility, can do charts of subscribers to the highest latency. One feature that is exciting is application latency. I can see a graph and a table that says your Netflix is a latency of X, YouTube is a latency of Y. You start to see how applications are behaving on your network. An invaluable feature for troubleshooting seeing help with the network. And, of course, you can do all sorts of metrics for the subscriber. The last piece out of all of this is the DOS (Denial of Service detection). While it is QoE there’s an added security feature that says I can now detect when a DOS attack is happening and then provide a list of participating IP addresses. We can at least see what IPs are violating or causing the DOS attack.
This is hardware and subscription-based correct?
Does the hardware sit at the tower site, or does it sit at your NOC (network operations center)?
We have both options. Let me elaborate on the hardware side a little bit as well as deployment models. First and foremost we don’t charge per subscriber subscription, it’s more of a speed base subscription so you buy a gigabyte worth of license that says you have an appliance that will traffic shape optimizes dpi rate limit up to a gigabit of throughput. It will go all the way from one to a 100GB solution. Very scalable. We sell a one-gig solution that we look at as a potential tower model. One might ask why are you doing a tower model? The interesting thing is we are coming across WISPS who are saying they don’t want to put everything through their NOC, but solve some of the problems at a particular tower with a low footprint and not spend a ton of money, it’s not a lot of traffic but it's a problematic tower with 500 users. They want to invest in the tower location. Plus they have a segregated network and an upstream provider, 5 towers that get fed by 1 and another 10 towers by somewhere else. That’s why the tower model helps. In parallel, you have the NOC model, and there you go from 1 gigabit all the way over to 100GB. In that model, we are also recommending that we give you the hardware specs and you buy your own hardware, or you can do a virtual solution. There is no point in Cambium going out and buying a Dell server and charging a premium on top of it. If you wanna by a Dell from Dell, buy it, meet the hardware requirements, we can give you the license and the server and you can go to town. That is how the model is. Yes, there is a subscription, it is invoiced quarterly and goes for as little as $270 a month for a 1GB to a couple thousand or more as you go higher up. Because it is a throughput-based solution the larger your network is the more subscribers you have, as long as you are below the throughput cap and keep adding more subscribers the economics will work out for you.
That’s great information. Earlier you were talking about WIFI, and we talk about WISPs all the time, would this be something that could assist MSPs also?
Bingo. Absolutely. In the world of Wi-Fi WAN optimizer application visibility where tens of hundreds of thousands of dollars a large company can go and afford that but for everybody beyond that scale, if you're an MSP doing an MDU or a small hospitality venue, you want to do things like traffic shaping, you want to create rate limit packages, and you want to optimize the protocol layer (TCP optimization). Anybody offering any type of broadband service can put this in and, benefit from it. MSPs are absolutely the right candidate because it gives them a much lower entry point from an investment perspective to go deploy, get the user statistics, and give the quality of experience. Wi-Fi matters. Devices that are old, new, old protocol, congestion, interference, etc. It does apply.
I see now where this could fit to help them, maybe a multi-dwelling unit where you have everybody roaming around with your apps and running crazy. So now when you say the subscription is based on the speed threshold would that be the speed and internet connection we are bringing into that area?
I think the best way to think about it is how much bandwidth your network consumes. Not necessarily the bandwidth that you’re paying for, you may be paying for a 100GB circuit, but in your worst-case scenario, you consume 3GB. So don’t go by 100GB license, buy a 3GB license. You only need to shape and manage the traffic that is actually going through.
Can you upgrade that service as you go?
Absolutely. The recommendation I would have is like if you are a 5GB customer buy the 10GB hardware. You’re not gonna pay that much more to go to 10 GB hardware, a higher SSD, a bit more memory, and a few more ports. Buy the 10 GB hardware so that if you buy the 5GB license and you have to upgrade you just buy the software license.
Where do you see this going, are there plans in the roadmap to enhance this more over the next year?
Absolutely. The future roadmap is quite exciting and I’ll give you some insight on that. You are going to see a lot more integration with C and CNmaestro. One thing that we now can bring to the table is you have your QoE box that is giving you things like application latency, Netflix latency, YouTube latency, etc. We get a ton of information on RSSI, SNR, our frame utilization, and packet error. Out of all of this, and even some things that our customers don’t see, we can get from CLI access under the hood. Envision that I’m gonna take all this application data, raw data, and collect those together and present a very compelling heat map or health history of the overall network. That type of statistics invisibility becomes a huge roadmap deliverable for us. On the QoE side itself you have time-based rules where your gonna apply rules at certain times of the day. And then there is a congestion detection and avoidance mechanism. Through artificial intelligence machine learning, you're going to watch the TCP congestion that could really be a congestion issue, I can then automatically apply some rules. Think automation. The appliance starts taking care of the network on your behalf so you don’t have to do much, that’s sort of the vision we’re gonna drive towards. There will be a lot of features along the way.
Looking forward to this. I’m already thinking in my head are some of the people that I’d love to set up calls and you know we can teach them about this.
The best part is trials. If you have the software base, try and buy is super easy. If you don't have a server we will ship you one with the software, run it for 45 days to see the results for yourselves.
Earlier you and I were talking about a feature that you guys are getting ready to release for ePMP, it doesn’t do everything the QoE does but it is a feature that’s going to be able to be purchased for ePMP, can you go to that a little bit?
We call this feature smart QoS. The gist of it is that in an ePMP 11 AC series of products with the 4.7 release (you can find it on our website) you can enable a DPI engine. You can do that for free. Basically, you can download 4.7, update all you’re APs and SMs, and right away start seeing what applications are going through your AP and SMs. It puts a nice little graph showing the high consumption and all that. Where we sell a license is to be able to take the application and map it to QOS (quality of service), not QoE, we all know the ePMPs, whether they are 450 APs. Wi-Fi APs a lot of times have these schedulers that say if it's a high priority, medium priority, little priority. Now with this dpi engine, you can go and say hey look on the web conferencing app that I just detected, I’m gonna bundle them together and I want you Mr. AP to put them in the highest priority queue. Or I’ve got some gaming customers behind this particular SM on that SM I’m gonna make gaming a higher priority in the uplink. The uplink is where things a lot of times matter because that’s where the bandwidth is limited. So this is what you can do. We did some tests inside the lab where we had high interference in an environment and we prioritize the zoom call and we saw a better performance on the zoom call with our QOS on vs off. Very attractive pricing, it is like $10 a month, $120 per AP, buy for a couple of years and let it run. Naturally, we know wireless will go to different cycles and we will have it again available on the E4000 for example. The same feature with AX, the exciting feature I think it’s gonna make a big difference for a lot of customers out there.
That is exciting. I think that being able to actually see what’s going on inside those pipes is just an eye-opener to where we can solve problems and find problems and make our networks run smarter, faster, and better. It’s great information, I really appreciate you taking the time to chat with us on this stuff. As always thank you for tuning into the Solution Series!