Poor results of network perfomance with 5.2.1-p9.

Nikolay Pavlov quetzal at roks.biz
Tue Aug 24 00:04:11 PDT 2004


Hi, FreeBSD users.

On Monday, 23 August 2004 at 14:10:36 -0300, Infomatik wrote:
> I am not so sure if your test is giving any usable result for FreeBSD.
> 
> It seems you are using some wireless equipment between both servers what means 
> that the throuput depends on this APs first.
> Any correctly cabled 10/100 NIC should give higher throughput than any 
> available 802.11b network equipment.
> Even if you have a clean radio connection (> 25Db SNR on each side) between 
> both APs, low(est) noise and a PP correctly configured you never get full 
> duplex , one direction ever is slower. Also it depends on the distance 
> between both points I believe. 
> There are also several issues with the wireless equipments, may be you never 
> get some usable cicles from an AP when it is highly used by another 
> connection already.
> 
> May be you first check your radio connection and before running a test between 
> both server you check pinging the remote AP (even flooding) to see how 
> capable your radio is. I guess you never get more than 1-1.5MB/s 
> bi-directional between both servers. You may get sustained transfer rates of 
> 3-6MB/s in one direction only but I do not know this APs enough to say it 
> exactly.
> 
> If you need higher traffic you should use Tsunami WL-PP-bridges or 802.11a/b/g 
> cards configured as adhoc instead of this cheap APs connected to your NICs

	You are right. This damned chinese AP again and again let down me. 
I have found a huge amount of drops on send queue in 5.2.1 side. When I have
drop speed down to 2 Mb/s on both AP results of test were leveled on a mark 
1.35 Mb/s. Sorry for that noise and thank's for your time.

Best regards,
	Nikolay Pavlov.


More information about the freebsd-performance mailing list