Freakishly high latency during file transfers

[CaroleSeaton]

With our Rogers Cable internet connection in Willowdale, Ontario, the ping time to the Toronto access point is usually 8 - 12 ms.  When attempting to use AFP to upload files from my Macintosh to another in Ajax (both running Snow Leopard) on a network router using the same tunnel server, the transfer speed is <160kbps and the IPv6 latency soars to 800-1200 ms. A Linux machine on my local (Airport Extreme) router reports the same high latency pinging the tunnel server over IPv4, which would seem to rule out a configuration problem on the Mac itself. Traceroute seems to indicate that the high latency appears at the tunnel server itself; the last hop on Rogers' network before the server has an ~8 ms latency. The latency drops back to normal the instant the file transfer is stopped even if the AFP connection is still open. Transfers in the other direction don't appear to be unusually slow although I haven't determined if any latency problems develop at the other end. Is this a problem with the router, with Hurricane, or some sort of "traffic-shaping" gone wrong?

[broquea]

160kbit/s or 160kbyte/s? If the latter, sounds like you are pushing the limit of the DSL in that particular direction. Since the IPv4 latency goes up as well, that would appear to be a congestion issue, and not related to the tunnel-server, otherwise IPv4 wouldn't be affected. you could try not going through the router or airport, and see if that makes any difference, but if that is your max upload speed, it would probably do the same again. I know I had one person open a trouble ticket recently and it ended up that their gateway appliance needed firmware update and they could get their full speeds again. However that wasn't a latency or congestion issue. It was more a rate-limiting issue going on in their gateway appliance, limiting each connection but they could open like 10 at all the same speeds.

[CaroleSeaton]

160kbit/s or 160kbyte/s?

It's 160kbit/s. (20kbyte/s)

If the latter, sounds like you are pushing the limit of the DSL in that particular direction.

The connection is 10,000 down/800 up.

Since the IPv4 latency goes up as well, that would appear to be a congestion issue, and not related to the tunnel-server, otherwise IPv4 wouldn't be affected.

Traceroute (not traceroute6) doesn't show any unusual latency for the intermediate hops on Rogers' network before the Toronto tunnel server. Shouldn't congestion affect every hop after the router?

However that wasn't a latency or congestion issue. It was more a rate-limiting issue going on in their gateway appliance, limiting each connection but they could open like 10 at all the same speeds.

The Airport router isn't set to throttle connections and there was no problem sending files in the other direction, only upstream.

[Ninho]

Call me paranoïd if you want, I'd suspect traffic shaping + deep packet inspection... ;=)