sometimes I have problems when loading some youtube videos.
To settle the obvious: when running these tests, I was the only one actively using the internet connection in the home and there was no torrent/uploading going on.
This happens only with some videos, and not with others.
I recorded a small video, demonstrating the problem:
Inside I show that:
- I have a stable 2Mbps internet connection (not very fast, but plenty for streaming a low quality video from youtube)
- The ping test towards youtube.com doesn't display any error
- Even with a 240p video, it stops playing after only 29s
- When pausing it to let it buffer, it apparently loads another 10 seconds, only to subsequently reset the buffering indicator to what is probably the true state
Again: this doesn't happen with other videos.
After a while, I realized that ping test doesn't really demonstrate anything: it's possible that youtube's CDN gives me a different server for different videos... and so I should ping test those servers.
I opened the chrome inspector, looked at
video/x-flv transfers and copied the hostnames for both that video and a video which loads vastly better.
r1.sn-nx5cvox-hpas.c.youtube.com is the bad one.
r1.sn-nx5cvox-hpal.c.youtube.com is the good one.
I tried to ping test these and this is the result:
--- r1.sn-nx5cvox-hpas.c.youtube.com ping statistics --- 59 packets transmitted, 53 received, 10% packet loss, time 274802ms rtt min/avg/max/mdev = 48.737/100.524/1060.337/189.638 ms, pipe 2 --- r1.sn-nx5cvox-hpal.c.youtube.com ping statistics --- 28 packets transmitted, 25 received, 10% packet loss, time 137016ms rtt min/avg/max/mdev = 47.244/97.993/672.556/140.503 ms
The results are quite worrying, but to me it doesn't explain the difference between the 2 servers.
By looking again at the youtube ping, I realize that, even if there was no packet loss, the
mdev was at 123ms... quite high indeed.
Is the difference in standard deviation enough to account for these problems?
I suspected that the internet connection could be subpar: for example, yesterday a
traceroute stackoverflow.com was failing after the seabone.
But I refuse to think that could be a problem common to all the people in my country... And in fact, by tethering with my mobile connection, I was able to reach stackoverflow.com just fine (fwiw: my landline isp is telecom, and my mobile isp is "telecom italia mobile"... So I'd assume that after the "last-mile" connection/the 3g cell connection, the route done by the packages should be the same... but apparently that's not true).
By trying to tether through the mobile connection, and stream that video, the problem still happens so: I cannot rule out the possibility that this is either a problem of the remote server, or of the internet connection.
Which is it?
If the latter: how common is it to have ping failings on only some servers, when using a consumer line?
If it's not normal behavior for a consumer line, could this be enough for requiring technical support? (not that I expect anything out of it: telecom is abysmal).
If it's a line problem: is it possible to avoid it, just like I bypassed the connection dropped at the seabone by using another connection? that is... is it possible to force the packets to another route to the target host? (I assume that resetting the router, and trying to get a different IP might help, but I never had any luck with that).