1 |
Am Wed, 21 Sep 2016 12:53:17 -0700 |
2 |
schrieb Grant <emailgrant@×××××.com>: |
3 |
|
4 |
> [...] |
5 |
> [...] |
6 |
> >> |
7 |
> >> You may want to set the default congestion control to fq-codel |
8 |
> >> (it's in the kernel) if you're using DSL links. This may help your |
9 |
> >> problem a little bit. It is most effective if you deploy traffic |
10 |
> >> shaping at the same time. There was once something like |
11 |
> >> wondershaper. Trick is to get the TCP queuing back inside your |
12 |
> >> router (that is where you deployed pppoe) as otherwise packets |
13 |
> >> will queue up in the modem (dsl modems use huge queues by |
14 |
> >> default). This works by lowering the uplink bandwith to 80-90% of |
15 |
> >> measured maximum upload (the excess bandwidth is for short bursts |
16 |
> >> of traffic). Traffic shaping now re-orders the packets. It should |
17 |
> >> send ACK and small packets first. This should solve your queuing |
18 |
> >> problem. |
19 |
> >> |
20 |
> >> Between each step check dslreports.com for bufferbloat. I'm |
21 |
> >> guessing it is currently way above 1000 ms while it should stay |
22 |
> >> below 20-50 ms for dsl. |
23 |
> >> |
24 |
> >> The fq-codel congestion control fights against buffer bloat. But it |
25 |
> >> can only effectively work if you're doing traffic shaping at least |
26 |
> >> on your uplink (downlink may or may not be worth the effort |
27 |
> >> depending on your use-case). |
28 |
> >> |
29 |
> >> Additionally, you can lower the priority of icmp-echo-reply this |
30 |
> >> way so during icmp flooding your uplink will still work. |
31 |
> >> |
32 |
> >> This link may help you: |
33 |
> >> https://www.bufferbloat.net/projects/codel/wiki/Cake/ |
34 |
> > |
35 |
> > And this: |
36 |
> > https://github.com/tohojo/sqm-scripts |
37 |
> |
38 |
> |
39 |
> I haven't mentioned it yet, but several times I've seen the website |
40 |
> perform fine all day until I browse to it myself and then all of a |
41 |
> sudden it's super slow for me and my third-party monitor. WTF??? |
42 |
|
43 |
I had a similar problems once when routing through a IPsec VPN tunnnel. |
44 |
I needed to reduce MTU in front of the tunnel to make it work |
45 |
correctly. But I think your problem is different. |
46 |
|
47 |
Does the http server backlog on the other side? Do you have performance |
48 |
graphs for other parts of the system to see them in relation? Maybe |
49 |
some router on the path doesn't work as expected. |
50 |
|
51 |
-- |
52 |
Regards, |
53 |
Kai |
54 |
|
55 |
Replies to list-only preferred. |