| 1 |
On Oct 23, 2011 7:07 PM, "Florian Philipp" <lists@×××××××××××.net> wrote: |
| 2 |
> |
| 3 |
> Am 20.10.2011 18:11, schrieb Florian Philipp: |
| 4 |
> > Am 20.10.2011 15:10, schrieb Pandu Poluan: |
| 5 |
> >> Like the subject said: I am wondering if using a non-default TCP |
| 6 |
> >> Advanced Congestion Control makes any difference. |
| 7 |
> >> |
| 8 |
> >> (The default is "cubic", but there are alternatives such as "htcp", |
| 9 |
> >> "hybla", and "yeah") |
| 10 |
> >> |
| 11 |
> >> Any experiences? |
| 12 |
> >> |
| 13 |
> > |
| 14 |
> > I tested it on the only situation I had where it was even remotely worth |
| 15 |
> > the effort to try it: NFS over TCP via an old and overutilized router: |
| 16 |
> > No measurable effect. I guess a web or mail server (read: something that |
| 17 |
> > is not primarily bandwidth constrained and where latency matters) might |
| 18 |
> > benefit more. But then again, how do you measure that reliably? |
| 19 |
> > |
| 20 |
> > You also have to consider where the client might be. A long distance, |
| 21 |
> > high bandwidth connection will benefit from different congestion control |
| 22 |
> > mechanisms than a local low bandwidth connection. |
| 23 |
> > |
| 24 |
> > Regards, |
| 25 |
> > Florian Philipp |
| 26 |
> > |
| 27 |
> |
| 28 |
> This paper and its references could be interesting. |
| 29 |
> http://research.google.com/pubs/archive/37486.pdf |
| 30 |
> |
| 31 |
|
| 32 |
Thanks! I'll sure to study them tomorrow. Productive time at the office ;-) |
| 33 |
|
| 34 |
Rgds, |
Archives