Bandwidth Test is not accurate if so please explain!

RouterOS General
1

Ok we have a 493AH and a 532 with two bonded Ethernet cables that are less than a foot in length factory made.

On both the 493AH and the 532 I have a bridge with the ports of the bonding and the Ethernet interface I am testing with.

So it looks likes this:

493AH
Bonding port ether1, ether9 mode balance rr link monitoring arp
interface bonding IP 65.78.123.50
bridge
ports:bonding1, ether2


532
Bonding port ether1,ether2 mode balance rr link monitoring arp
interface boning ip 65.78.123.51
bridge ports:bonding1, ether3



The bandwidth server is set up on .50 the 493AH
The bandwidth client is pointing to the server and this is the speeds I get with TCP both directions:
Tx/Rx 10s Averages = 18.8Mbps/18Mbps

With UDP I get with both directions:
Tx/Rx 10s Averages = 62.3kbps/141.8Mbps

Is the problem that I am doing the test on the boards their CPU goes up to 100% and stays pretty much there?

Shouldn’t a regular test with out the bonding show on a TCP both directional test at least 50Mbps?

Why do people always boast on posts of UDP data? UDP is the less used data on the network. Correct?

Please hit me back with some knowledge.

-Sincerely,
DesertAdmin

2

You asked for it:
http://en.wikipedia.org/wiki/Transmission_Control_Protocol
http://en.wikipedia.org/wiki/User_Datagram_Protocol

In simple terms - UDP is simple flooding other end, but TCP is waiting for ACKs for each and every packet.

and forget about any results if you generate traffic locally - you need to have 4 routers A—B—C—D, and run test from A to D (where C D are your routers)

BT are taking too much resourses to generate traffic, so you can’'t get any accurate results.