Slashdot Mirror
British ISPs Embracing Two-Tier Internet
Barence writes "Britain's leading ISPs are attempting to construct a two-tier internet, where websites and services that are willing to pay are thrust into the 'fast lane,' while those that don't are left fighting for scraps of bandwidth or even blocked outright. Asked directly whether ISP TalkTalk would be willing to cut off access completely to BBC iPlayer in favor of YouTube if the latter was prepared to sign a big enough cheque, TalkTalk's Andrew Heaney replied: 'We'd do a deal, and we'd look at YouTube and we'd look at BBC and we should have freedom to sign whatever deal works.' Britain's biggest ISP, BT, meanwhile says it 'absolutely could see situations in which some content or application providers might want to pay BT for a quality of service above best efforts.' PC Pro asks if it's the end of the net as we know it."
Clearly, you are misunderstanding the purpose of 21st century governments. The purpose of your government is to ensure that corporations and their shareholders become wealthier.
Palm trees and 8
"Best effort" in networking terminology is the priority given to traffic that isn't specifically prioritised or limited. There's nothing wrong with what he's saying.
No, it tells the user that their ISP is using Mafia tactics and saying "Pay us for speed protection or something unfortunate might happen".
"The tree of liberty must be refreshed from time to time with the blood of patriots and tyrants." ~Thomas Jefferson
I work for a company that is heavy involved (among other things) in just that sort of deep packet inspection technology. If you don't think that large ISPs are (or will shortly be) doing traffic shaping, you're a fool.
Except for the fact that it doesn't make any sense. How can it be the "best effort" if something can be prioritised ahead of it?
It makes perfect sense. When a QoS scheme is being designed; traffic is divided into classes, and (typically) each class is assigned to queues based on priority; each queue has a certain size.
The class that is not associated with any priority queue at all is called "best effort". The reason it is called best effort, is, unlike other traffic classes -- there is no priority or reservation.
Other traffic has priority in the form of something close to a guarantee; meaning, if prioritized traffic does not exceed the size of the priority queue, it is guaranteed to be delivered even in the face of congestion. Whereas the remaining traffic is just "best effort".
The traffic that is best effort will be delivered if possible (in the face of congestion), but it might be dropped, best effort is weaker than guaranteed priority.
queue 0 - Best Effort
queue 1 - Expedited Forwarding
queue 2 - Assured Forwarding
queue 3 - Network Control
The way priorities work is you generally reserve a small % for Network Control messages. Usually this is between 1% and 10% of the link, depending on how big it is (obviously you don't want to use 10% of a 10gig link). This is for stuff like your routing protocols, which HAS to get through or else nothing works.
Then you have three different priority queues. Queues 1 and 2 are reserved bandwidth queues, where you reserve a certain amount of bandwidth for traffic in those two priorities. Usually you'll have queue 2 reserved for a few select things like Voip or voice traffic which absolutely HAS to keep flowing. Queue 1 is usually used to manage other internal ISP services such as DHCP requests, although that's not always the case. Often this is where customer who pay for dedicated bandwidth have their traffic, although usually other shaping mechanisms are used to guarantee the bandwidth is available.
This leaves you with the 'Best-Effort' queue, called such because it makes its Best Effort to get all the remaining traffic pushed through with whatever bandwidth is left.
If you're operating a small network or an Enterprise network, as opposed to a Carrier-grade network, you usually don't need to mess around with it much. If you run SIP within your office, or have business applications which need a minimum bandwidth reservation, then you'll probably make use of one of the guaranteed bandwidth queues. A lot of larger businesses will use network control for network management, and then use queue 2 for things like SIP traffic, queue 1 for all business applications, and leave queue 0 for all the general internet traffic.
They might also implement it via RED. As an outbound queue fills packets going into that queue start getting dropped. This is done to prevent TCP global synchronization, and is standard practice. But if you change the rules a little, saying that packets coming from payer X get into the queue more often than non-payer Y, you've effectively lowered Y's performance during congestion without impacting Y during non-congestion.
Or it could be done via managing router queues. In order to route a packet you must inspect it anyways. Instead of having 1 outbound queue from a router you have two, three, or more. The outbound port sends from the high-priority queue more often, but still sends from the lower priority queues, or else to the hosts it appears to be congestion or a dead connection.
It could also be done via policy based routing on the AS level. An AS is a set of routers divided from other routers by political divisions rather than any technological reasons (AT&T routers vs Verizon routers). Each AS communicates routes and speeds via BGP. You direct the payer packets towards the faster AS you're connected to, then send the non-payer packets over to the slower AS.
There is no real way to speed up some packets without slowing down others, unless you literally build a whole new faster network, in which case why not put the other packets on there too?
That doesn't make any sense. If they are really making their best effort, it would be the same as guaranteed priority.
You're missing the point. These are accepted industry terms, not subject to dispute by people who think they know something about computer networking. Suggesting there is something wrong with the terms is just equivalent to brandishing your own errors/lack of understanding:
Quality of services clases impose constraints on the system, by assigning traffic to different queues with different sizes and latency/drop preferences. The priority queues are constraints, and not to be violated in order to deliver a packet. They are usually of a fixed size, so even some packets from prioritized streams can wind up dropped into lower priority classes.
Best effort refers to the best effort to deliver packets in that class, without violating any constraints imposed on the system.
Priority queues are constraints; packets in a higher priority queue cannot be dropped at a higher rate than allowed, for the purpose of delivering best effort packets.
Best effort is basically near the very bottom. The only worse traffic class is called Less than best effort; typically applied to non-drop non-latency sensitive traffic such as bulk transfers, e.g. FTP, FXP, SMB, Peer to Peer networks, sometimes SMTP.