So, as it relates to bandwidth requirements for a wireless internet service provider, is there a rule of thumb for calculating how much bandwidth you require for your subscribers?
Let’s use the example of a 30 subscribers with an average package of 5mb down and 2mb up.
Nowadays, the average home has:
-3-5 mobile phones connected passively uploading daily pictures to the cloud,
2 Smart TVs streaming Netflix
2 Computers browsing social media sites and youtube’ing
1 Console Game (i.e. xbox one / PS4)
Is this a case where you just have to start with a package and let it grow as bandwidth is depleted?
It’s ultimately a betting game. You can calculate your best and worst case scenarios, but the actual speeds that the average person will experience are ultimately a gamble of how many people there will be at the time the user tests their connection.
Your best case scenario “customers*speed=speed_you_need”, e.g. if you have 100 customers, 500M/200M is what you need to provide 5M/2M guaranteed speed for each customer.
Your worst case scenario is “speed_you_have/customers=speed_you_can_guarantee” so f.e. if you have a guaranteed 100M/100M speed available and have 100 customers, you can guarantee up to 1M/1M, and if you offer all customers 5M/2M, then certainly not everyone will be able to experience it all of the time.
You can calculate how many customers may experience their maximum speed at once (before any speed drops begin) by swapping the variables in the equation above, i.e. “speed_you_have/speed_you_promise=customers”… e.g. if you have 100M/100M and promise 5M/2M, then no more than 20 customers must be downloading at the same time in order for said 20 people to experience their maximum download speeds, and no more than 50 customers must be uploading at the same time in order for said 50 customers to experience their maximum upload speeds.
I have towers with 30 clients which use 60Mbps during peak hours. I have towers with 50 clients which use 12Mbps during peak hours. It depends on the customers. Younger clients tend to use more bandwidth than older clients. Wealthier clients tend to use more bandwidth than less wealthy clients. But there are no hard and fast rules. You don’t get to use good statistical averaging until you get to populations of many hundred clients. For us, it varies wildly from neighborhood to neighborhood. Overall, with approximately 2,000 clients we see about 750Mbps download and 110Mbps upload usage on the peak night of the month at the peak time. That is approximately 380Kbps/55Kbps per customer or 11.5Mbps/1.65Mbps for the “average” group of 30 clients. That won’t touch my heavier usage neighborhoods.
If you billing method is based on bytes transferred, users will typically be more careful about not leaving a TV streaming video while they are not in the room, which helps reduce your bandwidth requirements.
If you can start with a 100Mbps or 1Gbps physical connection with a small to medium commit rate, which can be quickly increased or decreased. You can get started and then size your monthly commitment to match your customers. Start high if you can afford it, just in case. You don’t want your network to be perceived as slow, if your customers have other options.
You can get away with less bandwidth if you do a lot of Quality of Service work on your network. QoS is an advanced topic. Bandwidth is often less expensive than the time, or hardware/service, to get the QoS right so that users are slowed down without perceiving the slowdown.
And the others friends are in the true. but you must do a SUM for your CIR customers, is the starting point. Remember, is very improbable to all customers navigating at maximum speed in the same moment. When you navigate, you press, download, watch. The streaming videos download and stop and the people watching. You need to prioritize traffic and limit some kinds of traffic like p2p. For example, in a 100 Mbps I choose maximum 10 Mbps for p2p. NO more
Given my pricey bandwidth upstream, I can see this topic changing to “How so I conserve bandwidth” and “How do I make my service seem fast”.
Web cache (squid) is one way. But this is only for httptraffic. Qos is also another important tool (i.e. Give priority to http and VoIP). Any other ‘tricks of the trade’ ?
Just make sure you charge enough to cover your costs. If you cannot charge enough to cover your costs and you have to go to heavy cacheing / QoS, perhaps the business model does not work.
More general comments - Are you sure just one site will give 100% coverage to target area(s) as from experience LOS to customers generally require several sites and this raises the outlay and running costs.
