You are not logged in.
I'd really appreciate some advice on this from you fine people, since my ISP doesn't have a clue at all and I'm kind of stumped as well.
I'm on a 1.5mb connection in my student halls which runs through a squid proxy. Since last Friday, the internet is fine in the morning, then horrendously slow from lunchtime until midnight, then at pretty much exactly midnight it speeds up again. According to speed tests, the issue appears to be latency - I'm located in Bristol and a test to Birmingham shows a ping time of 1800ms (but downstream of 1.5mbps). Pages in a browser seem to hang for about 2 minutes, but when it does resolve it usually loads them quite fast. However, this only affects browsing - Pidgin/Skype/Thunderbird all run fine. This doesn't seem to be a Linux issue either, I've tested it with my Windows XP netbook and same result.
I stumbled across a way to 'fix' it, by using an SSH tunnel to my VPS in the US - why would that solve this though? I was thinking that perhaps I bypass some of their network by using an SSH tunnel so its a network problem, but they say everything is absolutely fine. Unfortunately I can't ping/trace/mtr because this 'ISP' blocks them.
The weirdest thing here is the fact that it clears up at midnight on the dot, and I can get full speed. Any ideas on what this could be would be greatly appreciated.
Offline
I have both tcpdump and wireshark, but to be honest I don't know what I'm looking for - could you give some advice?
Offline
Sounds like they have configured squid to limit the bandwidth usage at night to accommodate a higher load of users.. Or they have setup traffic shaping to do the same thing.
Your SSH tunnel probably works because you are not going through their proxy anymore...
Just a guess, but my school throttles the bandwidth constantly with squid..
Offline
Thats what I thought at first, but the connection was fine for months and the ISP have said there is nothing that they know of that could be causing it.
Connection returned to normal at 11:20pm last night instead of midnight, still haven't got a clue as to what it is.
Offline
Squid has (or used to have) a problem once log files reached a certain size. Most crons run at around midnight, so I'm guessing that a squid -k logrotate clears up the issue.
The other possibility is that squid is running out of file descriptors. This error would be present in the cache.log. Again a squid -k reconfigure in a cron job would likely clear this up if not restarting the squid process entirely.
If this is an enrollment period, then the influx of new students could cause either one of the above issues.
Offline