1 |
Am 20.10.2011 18:11, schrieb Florian Philipp: |
2 |
> Am 20.10.2011 15:10, schrieb Pandu Poluan: |
3 |
>> Like the subject said: I am wondering if using a non-default TCP |
4 |
>> Advanced Congestion Control makes any difference. |
5 |
>> |
6 |
>> (The default is "cubic", but there are alternatives such as "htcp", |
7 |
>> "hybla", and "yeah") |
8 |
>> |
9 |
>> Any experiences? |
10 |
>> |
11 |
> |
12 |
> I tested it on the only situation I had where it was even remotely worth |
13 |
> the effort to try it: NFS over TCP via an old and overutilized router: |
14 |
> No measurable effect. I guess a web or mail server (read: something that |
15 |
> is not primarily bandwidth constrained and where latency matters) might |
16 |
> benefit more. But then again, how do you measure that reliably? |
17 |
> |
18 |
> You also have to consider where the client might be. A long distance, |
19 |
> high bandwidth connection will benefit from different congestion control |
20 |
> mechanisms than a local low bandwidth connection. |
21 |
> |
22 |
> Regards, |
23 |
> Florian Philipp |
24 |
> |
25 |
|
26 |
This paper and its references could be interesting. |
27 |
http://research.google.com/pubs/archive/37486.pdf |