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