Bell Canada dsl meddling?

May 20, 2008 12:12 PM

Can anybody tell me with some authority if it's possible to measure the effects of packet management/traffic management/deep packet inspection within the Bell Canada wholesale DSL network? The effects ought to be felt by any user whose ISP buys DSL service from Bell, but the question is really what effect the shaping has on normal things a user might try to do (e.g. peer-to-peer, but also other stuff).
posted by ashbury to technology (4 comments total)

I'm asking for somebody else, by the way.
posted by ashbury at 12:13 PM on May 20, 2008

You might be able to infer your traffic is being throttled or shaped with some load test tools, if you have an amenable target (one that doesn't mind being pounded by you), particularly if it will log connections, and show you logs to compare to your own. Packet inspection is only going to appear to you as increased latency, and with high performance routers, hardly any of that (so it is very difficult to discern).

There is no performance guarantee in the public Internet, however. So, just because you think you see some pattern of timing or packet latency, consistent with traffic types or network addresses, doesn't mean you can prove some kind of shaping is happening. On public networks, one man's "meddling" may be nothing more than the network doing load shedding to protect itself during unexpected traffic peaks (UDP floods and DDOS attacks are still pretty common behavior in the public Internet). Or, it could be something as innocuous as a local DSL concentrator or DSLAM that has been configured to give QOS priority on local loops for VoIP service bundles.

Even the telco wire plant and subscriber density can make for wildly varying DSL performance, at different times of day, since, as crosstalk from many users in a DSLAM wire bundle may increase at certain times of day, high speed DSL modems can react by doing something as simple as slowing down on excessive retrain errors. DSL, by nature, is a "best effort" service. You can infer very little with regard to traffic shaping being the cause of any performance variance, so long as residential DSL is your physical transport medium.
posted by paulsc at 12:56 PM on May 20, 2008 [1 favorite has favorites]

Yes it is possible, but you'd need to narrow your criteria significantly be defining your endpoints. You'd possibly be looking at identifying a host that sits just beyond the edge of the provider network to server as one of the endpoints and another host within the end user provider network, conceivably this is your DSL line or user. You'd then run any number of tests to determine what the traffic thresholds are.

Or you can go here, and use this, it's probably easier.

As to the authority, no idea what you're looking for, references? Azuereus wiki has resources, google has resources/references.
posted by iamabot at 1:00 PM on May 20, 2008

I can tell you they are definitely shaping voip and most udp to have a higher priority than http and p2p. If they didnt you'd be asking for an ISP that put your gaming and voip packets ahead of the guy who is downloading Bob Marley's discography all night. Not all shaping is bad.
posted by damn dirty ape at 2:03 PM on May 20, 2008