[prev in list] [next in list] [prev in thread] [next in thread] 

List:       voyage-linux
Subject:    Re: [Voyage-linux] Re: [Madwifi-devel] Real throughput tests at 4.4km
From:       Edwin Whitelaw <Edwin.Whitelaw () nrvunwired ! net>
Date:       2006-04-23 3:40:01
Message-ID: 444AF711.1050407 () nrvunwired ! net
[Download RAW message or body]

Just to add some data...

Two WRAPS with EMP8602s about 1 foot apart with the u.FL connectors 
facing each other.  -ng svn 1491 (voyage 0.2 upgraded to latest).  RSSI 
is in the -60db range.

Iperf to *or* through one of the WRAPS takes about 80% CPU @ ~26mbs. 
(running vmstat 1).  Neither the vmstat process or iperf itself seemed 
to contribute much to the CPU load. The CPU consumption is directly 
proportional to throughput so changing the rate to 24M (~16.7 mbs real) 
drops CPU to only ~50%.  An interesting sidenote on this is that I have 
a 266MHz PII inserted as a bridge in my home net and it shows 
substantially less CPU hit at the same throughput (iperf tests were 
between hosts on either side of the bridge).

It would appear that at max throughput, the WRAP does not have a lot 
left over for other work.  For 802.11b AP usage, it probably would not 
be very heavily loaded by throughput.

I did not see the asymmetric CPU usage or data rate that Beat is 
seeing.  It looks the same in both directions in his test.

Edwin

Beat Meier wrote:
> Wang wenjuan wrote:
>
>   
>> Good, and How about your CPU's frequency?
>> Thanks
>>
>>     
> Both sited same: Wrap 2c cpu freq 266Mhz
> I have tested the same stuff in lab and it's the same.
> If the iperf client is on the ap site the througput is much better!
>
> Beat
>
>   

-- 
<=+=+=+==+=+=+==+=+=+=+=+=+=+=+=>
Edwin Whitelaw, P.E.
New River Valley Unwired, LLC
2200 Lonesome Dove Dr
Christiansburg, VA 24073
540-239-0318


_______________________________________________
Voyage-linux mailing list
Voyage-linux@list.voyage.hk
http://list.voyage.hk/mailman/listinfo/voyage-linux
[prev in list] [next in list] [prev in thread] [next in thread]