I just finished upgrading a 2.8.28 to a 2.9.13 link between 2 buildings. System is a 1.8ghz Celeron on an Abit motherboard, CM9 wireless cards, -57 signal, and 1/4 mile distance.
I’m using 5.3ghz for the link since it’s short range (and I use 5.8 on my nearby tower). Using AES-CCM with WPA2 for encryption, I got 68Mbps TCP throughput while testing on the same unit (internal BTest). That’s NStreme, exact-size, 3200bytes.
Well, I know that it is probably possible to get more, however I am using 50 feet of LMR400 cable on each end, plus I found out that time I was using these cheaper pigtails (low gauge wire) that are worth a few DB of loss as well. I imagine that I’m losing 10-12 db in the whole deal, I have seen lower signal as well, sometimes as low as -50. The dishes are 28dbi parabolics with radome covers, another db of loss there. Not to mention the palm tree that’s in the fresnel zone (30 ft tower to 35 ft roof, across a supermarket parking lot with a tree and overhead parking lot lights)
I’m very happy with the performance, although I would like to get bonding to work correctly, I like the devil horns there - because it’s not just one of these links, there is actually 2 of these links in parallel, currently using dual-gateway routing tricks to balance between links. If bonding would work, which it didn’t in 2.9.13, I am guessing I might get 2 x 68Mbps TCP, which is 138Mbps. At least it’s got a complete gigabit network connected to it!
I’m scared to try bonding again - this is a link that trunks the telephone services between a city hall and the police station, plus provides the VPN connectivity for the patrol cars. That’s 24/7 - last time I tried bonding it hard locked up the router, and I had to reboot it in person. Scary.
True, but I can turn off one link without messing anything up and at least play with channel settings and such - like one link is 5ghz, normal, and the other is 5ghz-turbo, max NStreme optimizations, etc, so at least if that portion fails the other is still there as a standby.
We have a partially complete AP and backhaul site that was running 2.9.5. The site is set up w/a 14dBi 180 2.4ghz sector and a 29dBi 5.8ghz backhaul antenna. The backhaul is not yet running, but we had very marginal contact with the new site from another of our APs some 31km distant through another 14dBi 180 sector. At -92 dBm the signal from the new site was not sufficient to make it through the local noise on standard 802.11b channels to yield a useful connection.
We did an upgrade to 2.9.13 on the new site by changing to an intermediate and slightly quieter channel and uploading via ftp. After that our RSSI from the new site jumped from -92dBm to -86dBm and we’re now able to work it via the normal channel.
A remarkable performance considering the radio at the unfinished site is a CM9 w/no amplification and the antennae at the respective sites are both only 14dBi.
Unfortunately we still have to do the 40km drive from sea level to 880 meters in order to complete the alignment of the backhaul, but at least now we have a reliable back door into the site. Woohoo!
We’re also seeing vastly reduced CPU on nstreme compared w/2.9.5
Joke:
That is bad news … Now I should upgrade my main wirelesses to the new version . And I should choose I night without sleeping .
End Of Joke
Congratulations to MT for the good update. Keep up the good working !
Will there be problem with upgrading first “the core” to 2.9.14 and after that the other routers. I mean is there problem 2.9.14 to work with 2.9.8 - wds/nstream ?
Don’t know about WDS, but we’ve had no problems (so far knock on wood pray to appropriate cloudbeings etc.) w/one end running nstreme w/2.9.5 and the other 2.9.13 or 2.9.14