"Link tests" are done on RB itselfs using BTest Server / Bandwith test and UDP
"Real life test" are done like troughput traffic from our web servers to my personal computer using single or multitask downloads from FTP or WWW sever.
Ok i made some actual tests from outside of the link, so i cant give numbers from "real life test"
Only thing that is changed on the setup is the wirelless protocol from "any" to "nv2".
802.11 ..... 45,3Mbps CPU87/96
nv2 ......... 71,5Mbps CPU70/98-100 (TDMA period Size 2)
nv2 ......... 77,4Mbps CPU70/95-100 (TDMA period Size 3)
nv2 ......... 77,5Mbps CPU70/91-98 (TDMA period Size 4)
Changing TDMA-period-size makes some small difference in BTest and also in real troughput, but like you said "little bit"
I'll go home now and test the real troughput again.
So the real troughput
802.11 .... 38,5 CPU 79/100
nv2 ........ 21,2 CPU 29/63 (TDMA period Size 2-4, no difference)
looks like the 40Mbit is something like top of possible on the RB333, but the nv2 is no near to this barier, like the CPU is not on the 100load.
and the question again, what i am missing ?
And the second question is: will something drasticly change using 802.11N cards ? Is the problem just in the "old" R52 card and the R52N will solve it ? Yes i can change the RB333 for new RB433AH, but will it solve the problem with nv2 ?
I want to find whats wrong on this difference between test and real result. We dont have possibility to test real troughput on each AP remotly. And if i would see the +70Mbit test, i would expect +50Mbit of traffic, but there will be unexpected problem starting at 25Mbit load and this will end with unhappy customers.