Hi all,
In an rb800 with ros4.4. when I do a speed test to the routerboard’s own IP (i.e. no other routerboard, no cable, etc) I never see more than 106 Mbps (but it varies from 46 Mbps up to 106 Mbps) and the CPU is maxed out at 100%.
The rb800 is an 800 Mhz processor.
Is there a problem with ros 4.4 and the rb800’s ?
D
Just upgraded to ros 4.5 and no difference with the poor bandwidth test speeds (although now I see that the temperature is displaying correctly in System → Health)
D
Hmmm - maybe this is OK. I’ve just done the same bandwidth test (loopback really) on an rb493AH and its worse than on the rb800.
So I guess that the problem is either the bandwidth test itself (does it use a random data generator???) or maybe, for maximum throughput, the boards shouldn’t route but only bridge?
D
How exactly are you performing the test? If the board is not just forwarding the packets but terminates or creates the packets you’re not measuring routing performance, you’re measuring how fast the CPU can keep up with being an endpoint.
Hi Fewi,
I just got a shock when I did a bandwidth test but (accidentally) made the local rb800 the end-point for the test (I originally meant to test a wireless link with R52N cards). I saw speeds between 46 Mbps to 106 Mbps and then I realised that I was just doing an bandwidth test locally.
I wonder what eats up the processor? Is it the fact that the board is routing (rather than bridging) or is it just the bandwidth test itself that is processor hungry?
D
Generating the packets or receiving the packets and replying to them is expensive. Simply routing them between two endpoints connecting through the router is not.