I have some problems with devices that perform relaying role in multi-hop p2p network.
1) Relay router is having unusually high CPU load, peaking at 100% at relatively modest throughputs ~50Mbit (one way UDP test, 512 byte packet size).
2) Also I get really bad throughput results with 2 way bandwidth tests in such environment.
Basically in scenario where I have following topology:
R1 ~~~ R2 ~~~ R3
- links between routers are 5Ghz wireless 802.11n, 40hz channel width with WPA2.
- routers are 433AH with R52n cards, routerOS 5.5.
- Antennas used are cross-polarized panels, with 2 connectors and both chains set to active.
- using standard routerOS bandwidth testing tool on end devices (R1&R3)
I have no problems with R1 to R2 and R2 to R3 testing, even when devices themselves are running bandwidth probe and responder. All link characteristics are fine, signal close to -50dB, good CCQ, about 1km distance. For example R1 router running probe to R2 peaks at 100% CPU at around 130Mbit one way test. I am accounting that bandwidth probe is taking a fair share of CPU itself to generate all those packets.
However in scenario where R2 is relaying packets, with probe running from R1 to R3, I get bottleneck on R2 due to high CPU load, 100% at ~50Mbit. WPA2 doesn't seem to matter.
I suppose that something is wrong here?
To make things worse, there is another issue. When I run 2 way test (send+receive) I barely get 8mbit both ways. This time CPU load on R2 is not even close to peaking so I don't know where the problem is (and as before, simple p2p testing shows no problems at all, until I insert a device in between).
I'd really appreciate any help I can get with these issues
Hopefully I'll be able to assemble a lab for testing so I can check various tweaking on above scenarios, Nv2 etc, but for now I only have these semi-production routers in the fields where I am unwilling to make changes that can cut me off.