Thoughput - Link Capacity Test

Hello everybody.

I have some doubt about a little PMP430 network with 2 APs and all subscribers being SM10. The only network device for each SM is one IP camera to upload video images to the core, so the Download Data Rate is 20%.

In order to achieve an accurated Link Capacity test, I used Wireshark to capture network traffic by using Port Monitoring in the main switch. This traffic came from a fiber-linked camera, I had not the chance to use a wireless one. I think this should be ¿irrelevant? because I know MTU is 1500 for Ethernet and I guess it’s 1522 for Motorola air interface, according to the Link Capacity Test.

The fact is that Wireshark shows this percentage of uploaded packets from IP’s camera:

12% 80-159
10% 640-1279
77% 1280-2559

(The 100% of downloaded packets are between 40-79 bytes; almost each packet is a TCP ACK).

So the Uplink rate in the Link Capacity Test goes from 1.09 Mbps (73% Efficiency) to 8.26 Mbps (75% Efficiency) using these values.

Now, CactiEZ shows an average of 364.85 kbps outbound in the radio interface and maximum of 625.86 (cameras don’t support SNMP); these values are similar to those shown for the cabled cameras (455.92 avg, 715.01 max., for example), so I know measures are basically right and, eventually, the current used bandwidth is below the worst case in the Link Capacity Test, according to the Wireshark statistics for packet size.

So, when thinking aboout adding more cameras per SM, what packet size should I choose to calculate the real bandwidth they will use? I think that, by doubling the cameras per SM, I also double the chance for small packets to be scattered on the air, so the throughput would be “punished” and, eventually, I can not say that adding a second camera will use double bandwidth.

Did I miss something in this argument? What is really the Link Capacity Test measuring?

Thank you all.