Only Erol could explain this complex issue in such a straight forward way!!

The following is a re-working of several posts I made to ISP focused user forums. It was clear that there was much confusion and misunderstanding about issues relating to network congestion and contention ratios, and after several attempts to explain these issues I finally settled on the use of an analogy to get them across.

So, in the analogy imagine that the shared resource of ‘data pipe’ (whatever part of the network the pipe is in – first/last mile, middle mile or external connectivity) is a public toilet. An end user’s connection speed relates to how quickly they can ‘evacuate’ themselves. A 1mbps user takes half as long to ‘evacuate’ as a 512kbps user. Congestion occurs when you go to use the toilet and it is already occupied and thus you have to wait. A user’s data volumes relate to how much they evacuate.

At this simplest level then it becomes clear that as your ‘evacuation’ speed increases, the more people can share the toilet without experiencing congestion. This is because people spend less time ‘doing their doings’ and they are in and out of the toilet quicker, and thus the chance that the toilet will be occupied when you go to use it is lower. In network terms this means that the faster a user’s connection speed is the more users you can get to share a ‘pipe’ (i.e. the higher you can run that pipe’s contention ratios). This of course contains an implicit assumption that just because people can evacuate quicker they do not decide to evacuate more. In network terms people do, over time, tend to increase the amount they download (how much they evacuate) as their (evacuation) speed increases. However they generally do this slowly over time and as long as their increased usage is less that their increased speed then the above statements hold true. A dial up user going to 512kbps BB connection – approximately ten times increase in – will not suddenly download ten times as much data. Over time their data usage will increase but often to less than the times speed increase. Give them a 100mbps connection and this is even more true.

This toilet analogy can be used to look at other aspects of network contention ratios and network congestion. One idea people typically struggle to understand is the idea that a large pipe with a given contention ratio has less impact on users than a small pipe with the same contention ratio. Again the toilet analogy can be useful in explaining this concept.

Imagine a public toilet with a single bowl being shared by 20 people. This would give a TCR – toilet contention ratio – of 20:1. Now imagine a second public toilet with 20 bowls being shared by 400 people. This gives the same TCR of 20:1. In the first imaginary toilet it only takes a single individual that takes ages and ages to ‘evacuate’ to cause congestion. In the second imaginary toilet it would take 20 such ‘long time’ evacuators, all evacuating at the same time to create congestion. This is a lot less likely to happen than the first scenario. Thus in general terms the bigger the shared pipe / number of users sharing it, at a given contention ratio, the less impact it will have on users compared to a small pipe /number of users sharing it with the same contention ratio. This issue is particularly germane to comparisons between the two main BB technologies in the UK – DSL and cable. Cable tends to contend fewer users on smaller pipes than DSL and thus DSL can typically support higher contention ratios than cable with the same impact on users.

The toilet analogy is also useful at highlighting the issue of why many ISPs usage caps are flawed. For this example I will uses NTL’s usage cap as an example but the weakness in it is shared by many other ISPs that have usage caps. Again imagine a public toilet with 20 bowls shared by 400 people giving a TCR of 20:1. NTL currently define toilet abuse (network abuse) as being based on the amount people evacuate. However there may be a user that, whilst they evacuate 10 or even 100 times more than the 'average' user, they always do so between the hours of midnight and 5am in a massive evacuation session. Such a user never causes another person’s usage of the public toilet to be blocked - yet by NTL's definition they are a toilet abuser. This then is the most basic (but not only) flaw with the NTL cap as it currently exists. It is not how much you evacuate that causes congestion but when you do it.

Probably the most useful aspect of the toilet analogy is its use in destroying the myth, widely put about by ISPs that 3% of users account for 60% plus of congestion. Once you use the toilet analogy it becomes patently clear that even the heaviest of ‘evacuators’ – those that evacuate much more than the average – can only ever cause a maximum of one bowl’s worth of congestion (one user’s worth) and no more. Whenever any user is using a bowl and all the other bowls are also in use, then that user is causing congestion, along with all the others, regardless of how little or how much they may have used the toilet before and regardless of how much or how little they may have evacuated before. In fact not only can one user only cause one user’s worth of congestion – they can only cause less than one user’s worth of congestion and the greater the congestion the less that it is cause by a single users. This is because in a network, unlike a toilet, the ‘bowls’ are not indivisible single units. What happens in a network, when the 21st person comes to use the toilet and all 20 bowls are occupied, is not that the user has to wait for a bowl to be freed up. What actually happens is that the existing 20 users start using less than a full bowl, with the bit left over being given to the 21st user and so on with the 22nd and 23rd and following users. So the more congestion there is the less a single user creates it.

This may all seem like common sense, when put in the toilet analogy terms, but you would be forgiven for thinking otherwise given some of the rhetoric coming from some ISP’s on the issue of heavy users, and network abuse. This is especially pernicious if we are really sincere about building a BB Britain, because this metric – average usage – is the single most import metric. More important than % coverage and more important that % take up. The ‘vilification’ of heavy users undermines the idea of a BB Britain. Heavy users and heavy use should be encouraged and promoted, not condemned to be the subject of ISP propaganda. The real measure of how advanced a digital society we have built in the UK, compared to the rest of the world is how much higher our average usage is. So let’s stop the propaganda, let’s stop blaming heavy users for disproportionately creating congestion and let’s get on with driving up everyone’s usage and building BB Britain.