If I am trying to offer speeds from 100 Mbps to 300 Mbps, how many clients should I be putting on each of my access points? I am using the A5 and the A5c. Is there a good rule of tumb for this?
Depends on how much traffic you can push through the AP and what speed your clients are paying for.
I follow a 4/1 over subscription rule, we will oversubscribe each Mbps at most 4 times. It’s pretty simple and works well for smaller systems, various companies I have interacted with will have a different level of over subscription, I have seen between 2-40. Personally I think anything over 8 is foolish and will make your customers really unhappy.
An example of how this works out, let’s say I have an AP that will do 500 Mbps (and assume it’s the slowest point in my network, if you have anywhere else that is limiting your throughput then that is where you get your number). With that much speed, if you follow a 4/1 ratio, you would have 2000 Mbps of bandwidth you could sell. Which means you could have 20 customers at 100 Mbps or between 6-7 customers at 300 Mbps. You can do the math and it’s pretty easy to keep track of if you can “sell” more bandwidth or if you need to look at doing an upgrade.
Normally, I will start looking to upgrade a sector when it gets above 3/4 full. (So for our imaginary AP I would start planning on adding another AP when I had subscribed about 1500 Mbps on it)
If you get bigger, between 2000-5000 customers you can start calculating your over subscription by average peak Mbps per user. (right now most companies are between 3-5 Mbps) This relies more on statistics, but when you have large numbers of customers it works out really well and gives a more useful ruler of where you need to increase your bandwidth.
Thank you for your reply! So to clarify, when the access point specs say its max throughput is 1.0 gig does it mean that each client just gets a piece of that pie, or does the spectrum reuse change that?
So let’s say we have an AP that on the box says it can do 750 Mbps, what you will actually see will probably vary from that number.
Frequently that 750 Mbps is the aggregate throughput of the AP, so that would be if both Upload and Download are maxed out. If you are using a fixed frame technology (SRS in Mimosa) a portion of the airtime has to be set aside for Download and Upload even if there is very little traffic going in that direction. If you are using what is called a flexible frame or a non-synchronized system (normal 802.11) then the AP will dedicate more airtime to whichever it thinks needs it more (download or upload).
So if you had a Mimosa AP that was using SRS 50/50, at most you would see 375 (1/2 of 750) Mbps of traffic going in one or another direction. Mimosa is working on 75/25 right now so it should be hitting release in the next couple months. (75% of the airtime being dedicated to download because most customers only download stuff and upload isn’t very important for most use cases). You can use WiFi Interop, but then it’s much harder to reuse frequencies on the same tower or in the same area, but WiFi Interop will have lower latency numbers and be able to dedicate almost all the 750 Mbps to the download if there is that kind of demand.
Also, that 750 Mbps is frequently when the AP is set to it’s widest channel width. Wider channels do not reach as far and are more susceptible to noise. So if you are in a populated area with lots of interference or if you need to reach out to further clients you will want to run a smaller channel size. Smaller channel size means less throughput.
There are other limiting factors. That 750 Mbps is also the best case scenario, if your AP has to repeat itself to get the information to a client you are going to loose bandwidth to that. If you have a client with a poor connection, they will only be able to receive at the speed that their link to the AP will allow (so it could be you have a client that can only do 25 Mbps which would be really bad). If you have several clients that have bad connections the AP will have to spend so much time talking slow to them that it’s overall throughput will be lower and you wouldn’t be able to push 750 Mbps to any client even if they had a perfect connection.
If you had an AP with a perfect connection to a client then yes, they would be able to pull close to that that full 750 Mbps, if you had multiple clients connected to the AP with perfect connections then they would have to compete for resources if they both wanted to pull 750 Mbps at the same time.
There are some other “gotchas” as well. Which is why I recommend getting an estimate from what your AP is actually capable of doing. If you are like me and have lots of poor client connections, you have to sell lower speed plans because your APs just can’t go very fast. On the other hand, if you are really good about setting up your clients with only good connections, you will be able to push more bandwidth through your AP which means you can sell higher speed plans or have more customers connected to an AP then I can.
When offering internet through fixed wireless, is it typically an understanding that the customer gets what they get? Any idea on how those offerings work? What can a WISP guarantee?
@Noah, depends on the company and how hard they work on their customer experience/service.
To your question “is it typically an understanding that the customer gets what they get?” Sorta? If you install just any customer without regard to how good of an experience they will have off of your service, then ya what people get is what they get. But if you are selective to only install customers that you know you can provide a certain level of service to then you can control very closely how “good” of an experience your customers will get.
“any ideas how those offerings work?” Not sure what you mean, like shitty internet companies saying “you get what you get”? If you are looking for other types of bandwidth over subscription I talk a little bit about an option for you below, but you may want to check out some articles on the topic.
“What can a WISP guarantee?” See below:
When you are oversubscribing, you will always be playing games with the numbers, trying to keep people’s experience good enough that they don’t feel the need to speed test/you have enough free bandwidth available to do well on a speed test. Which really boils down to managing your network as well at setting clear service expectations, a fine balancing act between your sanity, profit and customer experience.
This is not a problem exclusive to Fixed Wireless: Fiber, Coax, DSL, Telephone, Power, heck just about any service works the same way. People subscribe for a specific speed/amount of service and you assume that they won’t use what they are subscribing for 24/7/365. That way you can lower your prices. They are happy because they get what they want at a reasonable price, internet access w/ the occasional high speed usage, and you can actually make money.
If you really wanted to, you can sell dedicated internet access, even off of a PTMP AP. You just sell service and bandwidth limit what each customer can pull so that at no point the AP can’t supply the requested bandwidth. I have a couple customers on systems like this. They pay a lot of money for dedicated internet access and I make sure they get their 20-100 Mbps at all times of the day. They pay an arm and a leg for it though, because it’s expensive for me to run. Your economics may vary, if you can make the math work out great for you, I can’t and I don’t even run the company I work for.
Does over subscription mean that you don’t deliver a good service? Absolutely not, consumer users average only a few Mbps at peak time. Netflix and YouTube use less then 6 Mbps for a 1080 stream. VOIP and Video messaging services don’t use more then 3 Mbps each (depending on the service and quality VOIP can be under 100k). Web browsing is less limited by your internet connection and more the hundreds of connections your computer has to make to get the web page. A good DNS server and smart bandwidth limiting can provide a pretty darn good experience on less then 10 Mbps. Video gaming is a big user of bandwidth, but that is only for game downloads/updates.
So if you are selling 100 Mbps connections you can probably get away with quite a bit of over subscription. I have a lower over subscription ratio because I sell lower speed service, largely because of my population base which is very rural and hard to get +50 Mbps to. With 100 Mbps plans you will probably want to look at doing a different system for over subscription, like leaving aside 120 Mbps of your bandwidth for people to speed test well on then sell off the rest of the bandwidth at the average usage you see at peak time. (5 Mbps per customer for example.) That said, you don’t need to just have 1 type of service you offer. Depending on how you do your bandwidth limiting you can not only have tiers of service, but also Quality of Service tiers where you can prioritize types of traffic and/or customers.
TLDR, you can guarantee and deliver pretty much whatever you want, it’s just a question of profitability.
That answers my question, Thank you!
Dear lord I hope so, sorry I can be pretty long winded sometimes, but it’s kind of a complicated topic and I felt like I could only gloss over so much info without muddying the waters.
Appreciate it, just trying to understand how fixed wireless differs from other internet offerings.
OSI Layer 1 technology, arguably layer 2 as well depending on what you are comparing it to.
Others should take careful note of the info in these posts - I know that I have - Thanks William5!
This is all really useful info including examples and real world concepts that illustrate how to use over subscription without killing your “first class” product you are offering.
Also, taking into account SRS mode and what it actually does to your throughput when in use…
Looking forward to seeing the 75/25 SRS mode option!