I've got MT 2.8.28 running on an epia via mini-itx board with 1ghz cpu (same hw both sides) with dual CM9's on a mini-pci expansion board. In the lab, I can acheive more than 1.70MB/s actual thruput end to end. In the field, the same hardware being 17 miles apart, I achevie about 320Kb/s actual thruput.
1) There is NO packet loss on the link. I'm talking repeated 100,000 1400 byte pings with ZERO loss (adaptive mode, which sends only as fast as the replies come back, effetively keeping the link near %100 full).
2) There is no latency on the link - ping rt (small 64 byte) is < 1ms all the time.
3) The built-in bandwidth test gives me a sustained 16mbps reading and higher (if I raise the rates). But connecting straight to it with the laptop and testing thuput directly to other side of link only yeilds 320Kbs and it does not improve if I lower it to 12mbps signalling rate.
4) Signal qualty is good - I have 68 about on both ends.
Details:
I have ether1 and an eoip interface bridged together. I have the nstreme interfaces numbered as 10.0.0.1 and 10.0.0.2 on either end. This was done to make a simple transparent bridge. CPU utilization appears to be about 6 or 7 percent while passing maximum traffic. In testing I had 2.9.2 here in the lab and nstreme-dual flat out did not workl, but loading 2.8.28 onto the boxes worked almost immediately.
The problem sounds almost exactly like what you should get with ack-timeout issues, but I thought that ack-timeout=dynamic took care of this?
The other observation is that the built in tester will give me very different results if I test bandwidth to the other side's eoip interface as opposed to the nstreme's ip. Testing to the nstreme is much faster than the eoip, when in reality, it's all the same box and I can't honestly see why there should be 5mbps or more just because I'm selecting a different interface on the same box. Here is an example:
[admin@] interface wireless nstreme-dual> /tool bandwidth-test 10.0.0.2 duration=20 protocol=tcp
status: done testing
duration: 21s
rx-current: 15.8Mbps
rx-10-second-average: 16Mbps
rx-total-average: 15.6Mbps
[admin@] interface wireless nstreme-dual> /tool bandwidth-test pub.lic.ip.addr duration=20 protocol=tcp
status: done testing
duration: 20s
rx-current: 11.3Mbps
rx-10-second-average: 10.9Mbps
rx-total-average: 10.6Mbps
Anyone have any ideas?
