| Quote from: Dex |
I must admit though, managing traffic during peak times is not acceptable, the infrastructure needs to be able to cope with the maximum demand. I work as a building management system engineer, there is absolutely no chance that our clients would accept that we could only meet the building design conditions during off peak periods......... we also deal with data centres, again, they would not accept that argument.
|
Well, there's management, and there's management. There's good management, and certainly scope for bad management.
For example, the telephone network is not designed so that everyone can make a call when they want - all trunk lines could be busy, or there's a vote in progress for some celebrities in a jungle; there is even a chance that you won't get a
dialtone when you pick up the phone - even wired phones. Why? Because the network is dimensioned for a reasonable peak-hour load, but not an excessive one... and the skills in dimensioning telephony equipment have been learnt over a century. If you want 100% guarantees, you will need to pay considerably more. I've worked for years as a telecomms design engineer, and I can guarantee that our clients need to be able to carefully dimension their equipment to meet peak load in the best way, but not to over-dimension. We also work very hard at handling the overload cases, so that the system continues to work in a coherent fashion, and continues to handle the peak load while shedding the overload.
Historically, datacomms has always been a much wilder place than telecomms - where the only rule was "best effort", and anything could be dropped at any time when overload occurred. There was no effort to keep any coherent order. But that is changing.
Traffic management is the means by which the datacomms world is learning to apply the century-old telecomms lessons to dimensioning. Applying management in the right way is good for you - it ensures that you get a fair share of the available resources at peak hours - and that you get that fair share in a way that tries to keep your applications (esp audio & video) working. When it is designed correctly, TM is fighting *for* you, not against you.
Why is it needed? It isn't *just* because too many people are browsing at once, or watching a lot of videos. One major reason is that there are protocols out there (especially P2P) which have been designed to use as much bandwidth as you can put in front of them - and so would overwhelm any ISP. Filesharers want as much bandwidth as possible, and don't really want to pay for it. In the old "best effort" world, these sort of protocols are very bad neighbours to anything else - particularly those activities that don't demand massive bandwidth, but do demand a reasonable quality stream, or a reasonable interactive response time.
If an ISP wants to be able to carry both types of traffic at their peak hour, unlimited, then they need to dimension their network with a lot of capacity - which, of course costs a lot of money, and puts up the subscription fees. And that capacity won't be used so well in the off-peak hours.
If, on the other hand, he wants to run an economically viable service, then he needs to purchase less peak capacity in the network (so his service is cheaper to subscribers), and persuade some of the traffic to time-shift into the off-peak hours (so it is well balanced). Having peak-hour usage limits and a policy of free overnight usage are both a stick and carrot that helps persuade the "bad neighbour" users to time-shift their non-interactive loads.
Traffic management provides two facets to help during the peak hours: It actively helps those applications that have interactive or timing needs (browsing, audio, and video), and it polices against applications that are trying to guzzle bandwidth. The downside is that the ISP has to make assumptions about how you want to work - and those assumptions don't always fit.
Any "good neighbours" who cannot time-shift their traffic are usually willing to pay slightly higher subscriptions, so that they can get their traffic through. Likewise those subscribers whose needs are different from the norm are also more willing to pay higher prices. And for those people, the packages do exist, as higher price points. You can choose to get your traffic through faster, and you can choose to buy more peak-hour traffic capacity.
Traffic Management has one more facet. Those who have learnt to use it properly will also tend to learn how to use it in times of abnormally high usage, or severe network damage. Traffic Management can come in very handy at continuing to give fair access when really heavily overloaded, or badly degraded (think of the impact to traffic of another 9/11, or an explosion at one of the Telehouse centres) - to those who know how to use it.
I agree that traffic management can be bad. It can be hidden, undocumented, and can be used to paper over the cracks of a badly under-dimensioned network by someone out to ripoff the subscribers.
The trick for an ISP is getting the balance right. The trick for a subscriber is identifying the ISPs doing good from those doing bad.
I just went looking for an old guide to the economics that I know is on the Plusnet blog somewhere, when I came across these 2 videos explaining what I just said:
Video 1of Dave TomlinsonVideo 2 of Dave TomlinsonThe
old guide to the economics of buying capacity on the internet
In fact, Plusnet have a huge amount of stuff on their blog that
justifies their position in traffic management. The stuff back in 2008, as the iPlayer started service is eye-opening.
Plusnet Customer
Using FTTC since 2011. Currently on 80/20 Unlimited Fibre Extra.
