| 1 |
Am Wed, 21 Sep 2016 07:30:40 -0700 |
| 2 |
schrieb Grant <emailgrant@×××××.com>: |
| 3 |
|
| 4 |
> [...] |
| 5 |
> [...] |
| 6 |
> >> |
| 7 |
> >> |
| 8 |
> >> I just remembered that our AT&T modem/router does not respond to |
| 9 |
> >> pings. My solution is to move PPPoE off of that device and onto my |
| 10 |
> >> Gentoo router so that pings pass through the AT&T device to the |
| 11 |
> >> Gentoo router but I haven't done that yet as I want to be on-site |
| 12 |
> >> for it. Could that behavior somehow be contributing to this |
| 13 |
> >> problem? There does seem to be a clear correlation between user |
| 14 |
> >> activity at that location and the bad server behavior. |
| 15 |
> > |
| 16 |
> > If that device behaves badly in router mode by blocking just all |
| 17 |
> > icmp traffic instead of only icmp-echo-req, this is a good idea. |
| 18 |
> > You may want to bug AT&T about this problem then. It should really |
| 19 |
> > not block related icmp traffic. |
| 20 |
> |
| 21 |
> |
| 22 |
> Hi Kai, yesterday I switched my Gentoo router over to handling PPPoE |
| 23 |
> and pings seem to be working properly now. The AT&T device is now |
| 24 |
> functioning as a modem only and passing everything through. Today |
| 25 |
> I'll find out if it helps with TCP Queuing and (supposedly) related |
| 26 |
> http response slowdowns. |
| 27 |
|
| 28 |
You may want to set the default congestion control to fq-codel (it's in |
| 29 |
the kernel) if you're using DSL links. This may help your problem a |
| 30 |
little bit. It is most effective if you deploy traffic shaping at the |
| 31 |
same time. There was once something like wondershaper. Trick is to get |
| 32 |
the TCP queuing back inside your router (that is where you deployed |
| 33 |
pppoe) as otherwise packets will queue up in the modem (dsl modems use |
| 34 |
huge queues by default). This works by lowering the uplink bandwith to |
| 35 |
80-90% of measured maximum upload (the excess bandwidth is for short |
| 36 |
bursts of traffic). Traffic shaping now re-orders the packets. It |
| 37 |
should send ACK and small packets first. This should solve your |
| 38 |
queuing problem. |
| 39 |
|
| 40 |
Between each step check dslreports.com for bufferbloat. I'm guessing it |
| 41 |
is currently way above 1000 ms while it should stay below 20-50 ms for |
| 42 |
dsl. |
| 43 |
|
| 44 |
The fq-codel congestion control fights against buffer bloat. But it can |
| 45 |
only effectively work if you're doing traffic shaping at least on your |
| 46 |
uplink (downlink may or may not be worth the effort depending on your |
| 47 |
use-case). |
| 48 |
|
| 49 |
Additionally, you can lower the priority of icmp-echo-reply this way so |
| 50 |
during icmp flooding your uplink will still work. |
| 51 |
|
| 52 |
This link may help you: |
| 53 |
https://www.bufferbloat.net/projects/codel/wiki/Cake/ |
| 54 |
|
| 55 |
-- |
| 56 |
Regards, |
| 57 |
Kai |
| 58 |
|
| 59 |
Replies to list-only preferred. |
Archives