Hacker News new | ask | show | jobs
by emj 3193 days ago
I get 60ms RTT for the largest bin of packets with WebUDP many requests goes as high as 140ms (positive skew of the distribution), while ICMP never goes above 29ms RTT. I know UDP can be badly prioritized in some networks but I don't think it can only be that...

This needs more investigations imho, I wonder where this fails so badly; is WebRTC aiming at ~100ms latency?
1 comments
lxtx 3193 days ago
Definitely a bit weird. Where are you located? The echo demo server is hosted in the Netherlands. That's how the round trip histogram looks for me in Estonia: https://i.imgur.com/iBoqwe4.png I'll take a look later if there's any noticeable delays caused by the server.
link
emj 3193 days ago
I left it running for some hours and got alot more 30ms responses:

https://imgur.com/a/SdWtc

link
emj 3193 days ago
Chrome use ~100% cpu transceiving ~83 packets per second, and I think it sends packets at a slower rate later on, that's why we get so many 30ms responses when we wait some hours.

This is a tcpdump after loading the page:

  \time sudo tcpdump -c 1000 -ni eth0 udp and port 9777
  1000 packets captured
  1000 packets received by filter
  0 packets dropped by kernel
  0.02user 0.05system 0:12.45elapsed 0%CPU (0avgtext+0avgdata 5472maxresident)k
  0inputs+8outputs (0major+1508minor)pagefaults 0swaps
This is just a flood ping, interesting that the UDP version shoots more packets.

  \time sudo ping -f -c 1000 188.166.46.128
  PING 188.166.46.128 (188.166.46.128) 56(84) bytes of data.
  ....    
  --- 188.166.46.128 ping statistics ---
  1000 packets transmitted, 996 received, 0% packet loss, time 14195ms
  rtt min/avg/max/mdev = 28.839/38.641/50.433/5.152 ms, pipe 5, ipg/ewma 14.209/41.787 ms
  0.02user 0.13system 0:14.31elapsed 1%CPU (0avgtext+0avgdata 2248maxresident)k
  0inputs+8outputs (0major+1183minor)pagefaults 0swaps
link
emj 3193 days ago
https://imgur.com/a/9KHYj

this is after start up.

link