I have for some time, tried to figue out a wierd thing about bandwith test to devices. I have no clue what so ever why this thing happening. Maybe somone have a tip the situation:


Router - wireless - wireless - router

R1: RB1200 - MPLS/OSPF/BGP
W1: Rb800 nv2 - N mode
W2: SXT
R2: Rb 750 - MPLS/OSPF/BGP

R1 is connecting to R2 via "linknet" (lets say 10.0.0.1/29)
this is distributed by ospf to the rest of net.

Both routers have loopback ip, lets say: R1 = 172.16.0.1(/32) R2 = 172.16.0.2(/32)
This is also distibuted by OSPF

my problem is now:

Bandwith test R1 - R2

R1-R2 (loopback ip) UDP Send = OK (about 90 Mbps)
R1-R2 (loopback ip) UDP Recive = OK (about 90 Mbps)
R1-R2 (loopback ip) TCP Send = Not OK (about 2-3 Mbps)
R1-R2 (loopback ip) TCP Recive = OK (about 50Mbps)

the strange thing is: If trying R1-R2 TCP Send, but to link address its ok, and about 50 Mbps. Why this different?

i was wondering this same thing last week... not sure if it has something to do with the bridge interface or what. would be nice to figure it out.

I found a similar issue regarding the loopback.
If you ping the loopback and nearest ip on the unit where the loopback resides, the ping times are higher on the loopback ip than the interface ip. Especially noticeable when there is a slower link speed (like wireless) between you (it's like the ping times double).
Might be a conincidence, but this was found on an MPLS network where only loopbacks are advertised via LDP.

I just tried the ping test and my ping times were identical between an interface IP and Loopback IP, averaged 6ms.

As far as TCP/IP bandwidth tests being un-accurate watch your CPU meter on the routerboards. They will peak out at 100%. I think the processors are not big enough to self generate that much traffic?