I don't like network transfer quotas or caps. I don't like them a single bit. On a wireless service they make some amount of sense, although historic behavior would indicate to me that if you're allowed to use a certain amount of a service in a month, you're either going to power through it at the beginning or the end of the month. This is a problem because transfer quotas are meant to reduce congestion on a network, and are also often talked about as helping mitigate the costs of moving data from one place to the next on a network. If congestion needs to be reduced on a network, lower the transfer speeds each client can attain, or if a network upgrade needs to be funded, increasing rates for all of the customers is a fairer way to do it.
As a micro-cosmic example, let's say that you have a network. Let's presume that this network is a bit bigger than a home, but a bit smaller than a whole town. A school of some sort is a convenient size. When you build your network, you put a switch in a fairly central place, such as in the office or the library, and you have a server in that place as well, and then you place client computers in various locations, such as teacher offices, classrooms, and a computer lab. For simplicity's sake, let's presume that there isn't an Internet connection. This network consists purely of services being delivered from a central server to client computers, and information transferred directly from one client to the next.
Once you put your server, switch, and wiring, there's not much else that's really involved in running or maintaining the network. Because all of this stuff consumes a relatively stable amount of electricity, your costs for running the network are the same whether a single client computer connects or a hundred of them do, and your costs are the same whether each computer transfers one megabyte of information, or one hundred gigabytes of it.
Of course, your network and server need to be beefed up a little bit if you're counting on a hundred gigabytes of usage from each client every day. It's pretty easy to allocate these costs if you're pretty sure that each classroom is going to use about the same amount of data as every other classroom. What if, however, the sixth grade classroom uses a hundred gigabytes of network throughput in some period of time, and the Kindergarten classroom only uses one or two gigabytes in that time? I think the question then is whether or not the network usage of the sixth grade classroom affects the kindergarten classroom. Is the reason the Kindergarten classroom uses so little network throughput because the sixth graders are pushing big video files all day, thus causing network congestion, or is the reason that they're kindergartners, and not really capable of using that much data. Maybe a better question is whether or not those high-use sixth graders are causing any problems for nearby groups, such as fifth graders, or if there are any, seventh graders.
In an environment like this, the solution is pretty clear – you can't typically impose a throughput quota on a LAN, nor can you start charging the sixth graders more money to use the LAN. Your only option if those sixth graders' video projects are clogging the network is to accept that clogging, or to upgrade or change the network so that it works smoothly when the sixth graders are working on their video projects. Maybe the sixth grade classroom needs to be given a faster link and be given a dedicated switch, or if you built your LAN in the '90s, you need to upgrade from a hub to a switch. You may also need to add more connectivity to the server, or give the sixth graders a dedicated server for their video.
What do you do if, say, instead of a network of classrooms, you've wired up an apartment complex and given each unit a computer? At that point, you're probably collecting some kind of fee for the use of the network, and your budget for upgrades to the network comes from those fees. To give our network somewhere to go, let's just pretend that you've started with a 10 megabit hub, or a series of hubs connected to a switch.
If everybody uses the network about equally, then it's not obvious that there's a single person you could ask to pay more for their use of it. If, however, there is a population of heavy users, it's easy to see why asking those users for more money might be attractive, especially if, say, one person's usage is impeding on the ability for people near them to effectively using the network. (The person living in Unit 17 shuttles video files back and forth on the network, and the people sharing a hub with them notice that their ability to receive e-mail and open text documents is slowed when those video files are flying across the network is reduced.)
At that point, especially if you (as the operator of the network) have a way to monitor each unit's usage, might be tempted to either impose a total transfer quota, or start charging more for usage beyond a certain level. You might even talk to the person in Unit 17 and say that he's really paying for document editing service, but he's editing video and he should upgrade to the video editing service.
The question in my mind is what you do for both Unit 17 and the rest of the network around that subscriber if you start charging more for "video editor" class service. Are you going to connect his unit directly to the backbone switch that has the server he uses on it? Will you upgrade the group of units he lives in to a ten megabit switch or a 100-megabit switch so that he gets what he needs, the subscribers around him get what they need, and the subscribers in the rest of the network haven't noticed anything anyway, or are you just going to pocket the change and continue either leaving the network as it had been, or along the same "upgrade everything at once when we have the budget" path you'd been on before?
In the real world, there's no such thing as a video editing package, and on systems where you can control the throughput of each port on a large switch (DSL and cable systems), it makes sense to sell different speeds based on how much the customer asks for. Unfortunately, "business class" rarely makes sense for somebody who is a big user, but isn't otherwise isn't using any added functionality, such as static IPs or running their own server. In our example with the video editor at unit 17, switching to tiered service speeds (by laying a DSL network and using a VDSL2 DSLAM to sell the video editor 40/20 service and 10/2 service to everybody else) may have been a reasonable way to both upgrade or change the network, move to a switched architecture so everybody gets what they need without affecting other customers, and find a way to either manage the users (by only giving each of them access to a limited amount of the total throughput) or have them pay more for access to the network with higher performance. Rate-limiting may have been effective from the start, as well. If Unit 17 didn't want to pay more, they can still make the most of their 10/2 DSL line without affecting the other customers, but their own capability is limited.
Of course, if you don't want to rearrange your whole network and only a single user is asking for more speed, or using so much throughput that the ability of the other users to use the network is reduced, then providing a better connection to our single subscriber may also be a good option. Offer to wire their unit for a 100 megabit switched connection that terminates closer to the network backbone. This will have a lower impact on the other subscribers and will the fees from this advanced type of service should pay for it.
If the subscriber at Unit 17 declines to buy faster service or have a dedicated line of some sort installed, then it may be time to deploy the management functionality of the network infrastructure. Reducing the available speed to each port to six or seven megabits across the board should make it so your other customers get the performance they need out of shared infrastructure that's only capable of ten megabits in total, and the video editor in unit 17 doesn't need to pay the fees on new infrastructure.
I'm unconvinced that doing something like imposing a limit on the amount of transfer will do anything to modify Unit 17's behavior, or help the other customers. None of the other customers were in danger of running up against the transfer limit, but nobody feels comfortable knowing that it's there, and right up until they run up the limit, Unit 17 is going to continue using their connection at full effort, thus still impeding on their neighbors' ability to use the connection anyway. This applies whether they use their quota at the beginning of the period, or wait until the end. (This is true both from a behavioral standpoint, that a quota isn't going to change somebody's usage patterns, just make it so they can only do them for a limited amount of time during the period, and many people don't have enough control over their individual computers to limit their transfer rate. (Plus, who is going to buy into ten megabit service, and then only use two or three of those ten megabits by their own choice?)
Recently, the tech press has talked about a few rural broadband Internet Service Providers whose actual situations isn't too far away from these. There's a provider in Iowa that charges DSL customers a fairly significant amount for the ability to transfer more data on their connection. The provider has a limited number of customers and they offer fairly slow service, but if you need to transfer about a hundred gigabytes in a month (not actually that difficult, even without putting full effort into it) then your bill will be around $300. Anything above that is considered pretty excessive in usage and is billed accordingly.
My question is what the story is here. Is it truly more expensive to run fifteen gigabytes over a data pipe than ten gigabytes, or is it simply a way to grab more money from whomever is willing to pay it for upgrades to the whole network? An Internet network is more complicated than a big LAN, is there some cost that the telco in question has now incurred that isn't immediately obvious to its customers?
In the example above about the video editor in Unit 17, one of the options was to install a switch to the network segment that serves that customer. The other customers benefit based on Unit 17 paying more for the type of work that they do on the system.
I think that if you have a case of a real world network service provider where one customer transferring a few extra gigabyte truly incurs additional costs, then the provider needs to work on upgrading their infrastructure. There's no real reason why in a DSL system, one customer's use should affect the others, except in a situation where the provider is over-promising the whole system, or there's simply not enough backhaul. (This is especially true in a case where the DSL customers want to communicate to a larger network, such as the Internet.)
For a while, I was attempting to prove this point to the Telco (mine, not the low-density Midwestern one discussed above) by transferring data from my work computer (where I have a 100M/100M connection to the Internet, which regularly speed-tests at about 96/96 or so) to my home computer at full effort. Hilariously, I truly believe that the only one affected by this was myself. My Internet connection at home is so slow that it's a drop in the ocean of the backhaul infrastructure of The Telco, and all indications suggest that they're either reasonably significantly under-promising/selling their overall network, or the path from my home to their backhaul is about as dedicated as it gets.
What's hilarious about this is that my Telco is deploying transfer quotas in my area. They're only enforcing them in one direction (download), which is even odder, and the oddest part for me is that they're not deploying the transfer quotas in every area. The Telco is made up of several acquired and merged companies, and they haven't yet unified or updated the infrastructure of the whole network to match, so the quotas are only deployed and applicable in areas where the network supports it, where I happen to live. The Telco as they exist in a few other locations, however, doesn't meter their connections at all.
There are two problems with this, of course: The first is that I think it's unfair to their customers paying the same amount for the same service from the same company in different physical areas to have different restrictions on that service. The second is that the transfer limits are deployed in a really weird way and it's nearly impossible to tell when you will be affected by the quota.
The quotas are measured on a three-monthly basis, but presented on a monthly basis, so if you use a third of your available monthly quota one month, but six times your quota (presuming your connection can do this) then you're warned for the whole period. Near as I can tell, they don't tell you what specific usage put you over. After one or two times, their solution is to sell you a business-class account, which features no usage quotas, but are on the same infrastructure (and use the same equipment) that the consumer lines do, so if your usage had been affecting your neighbors (possible if you and a bunch of other subscribers, all provisioned at 40m/20m all share the same gigabit backhaul line) it will still do so, but The Telco will stop bothering you about it and have officially lifted your limit, and may even have raised your priority in their switches, at the further expense of their other customers.
Unfortunately for me, The Telco doesn't talk very much about their plans to upgrade their infrastructure, so there's no real indication as to whether I should switch providers now, or wait for the upgrade of their common infrastructure, with or without paying more for business-class service. (In fact, there's no indication as to whether my paying for business class, or adding more services like telephone will help get the service upgraded.) In addition to not having much information about their upgrade plans, the company seems to be totally unwilling to sell the dedicated fiber lines it advertises as a Metro Ethernet product. Nobody I've been able to contact has had any information as to whether or not it would be easy or even "possible" to install the Metro Ethernet product here, even if I do pay for it. (Similar to just giving Unit 17 a dedicated line in the earlier example.)
For a society as interested in collecting money as ours is, and for as much as people like to proclaim that you can get anything done with money, I've been finding that it's difficult to throw money at some types of enterprises in order to get additional service. I understand that it's easier when you've wired up just 100 dwelling units, recently, and completely privately, but I fail to see why this type of documentation and knowledge about the local network doesn't scale up when you move from a single landlord's property to a city managed by a team.