1 |
Can't sleep much last night due to the baby's quirky sleep pattern, so |
2 |
been spending the night reading about TCP congestion avoidance |
3 |
algorithms [1] -- and bufferbloat [2]. |
4 |
|
5 |
So, here's a spark of what (I hope) could be a starter for a deep |
6 |
discussion for us sysadmins/netadmins. (Which is why I decided to post |
7 |
in the -Server list). |
8 |
|
9 |
I've been, um, enjoying bandwidth beyond what I should be |
10 |
experiencing, by extensively tuning the TCP buffers (among others, |
11 |
increasing its max-size to, like, um, 16MB) and using a non-default |
12 |
congestion avoidance algo (namely, H-TCP). However, reading the |
13 |
article from Jim Getty about bufferbloat made me re-think my setup. |
14 |
|
15 |
Another factor that made me re-think my setup is the 'strange' |
16 |
characteristics of traffic between my office and our |
17 |
brand-spankin'-new subsidiary office 14 floors below us: SSH is very |
18 |
nice, but any big file transfers (sftp, http, ftp, cifs, *anything* |
19 |
biggish) will run well only for the first 10 seconds or so, before |
20 |
slowing to a crawl (and even managed to make WinSCP complaining of 'no |
21 |
response for 15 seconds'). But the ping's have no dropped packets at |
22 |
all. |
23 |
|
24 |
Oh, and this slowdownishness also affects some app in the sub office |
25 |
which connects to the MS SQL Server @ HQ. |
26 |
|
27 |
I'm in a bit of quandary; my bandwidth may be ill-gotten, but |
28 |
employees (and the Management) enjoy the current performance. But, |
29 |
then again, connection between HQ and Sub is flakey. |
30 |
|
31 |
Reading more on the comments given in [2], I had thought that traffic |
32 |
shaping -- at least between HQ and Sub -- may be the solution. And I |
33 |
might alleviate the interconnection pressure somewhat by changing to a |
34 |
different algo. |
35 |
|
36 |
Unfortunately, howto's and guides re: traffic shaping and TCP |
37 |
cong-algo that I can find using GooBing seems... outdated. |
38 |
|
39 |
So, some questions popped into my head: |
40 |
|
41 |
1. Is there a quite-recent and near-up-to-date guide to traffic |
42 |
shaping and/or TCP cong-algo? |
43 |
|
44 |
1a. Any analisis on the pro's/con's of H-TCP vs CUBIC vs Westwood+ vs |
45 |
(insert some exotic-but-available cong-algo here)? |
46 |
|
47 |
2. Am I experiencing and/or causing and/or contributing bufferbloat? |
48 |
|
49 |
3. What else could explain the peculiar traffic characteristics |
50 |
between my HQ and my sub-office? |
51 |
|
52 |
3a. Will egress traffic shaping help? |
53 |
|
54 |
Any thoughts are welcome. This thread is open for discussion. |
55 |
|
56 |
And forgive me if I write somewhat incoherently; I'm yet to finish my |
57 |
first cup of coffee. |
58 |
|
59 |
[1] http://www.google.com/m?client=ms-opera-mini&channel=new&q=linux+tcp+congestion+avoidance+algorithm&spell=1&ei=wxYNTpjZM6eViAKkpfSCAg&ved=0CAQQvwUoAA |
60 |
|
61 |
[2] http://gettys.wordpress.com/2010/12/06/whose-house-is-of-glasse-must-not-throw-stones-at-another/ |
62 |
|
63 |
Rgds, |
64 |
|
65 |
|
66 |
-- |
67 |
-- |
68 |
Pandu E Poluan - IT Optimizer |
69 |
My website: http://pandu.poluan.info/ |