a week ago - last edited Sunday
I installed 8.3.0 on an existing Rocket 5AC Prism AP with 7 AC clients. Without doing anything else, download speeds went up from about 45 to about 60 megabit. However, switching to pure AC mode dropped capacities to about 20% of what they were in mixed mode. Speed tests from the AP to the client radios supported this precipitous drop in capacity.
I didn't fiddle with it much, since the AP is in use, but I tried 5ms and 8ms fixed frame with 50/50 and 67/33 duty cycles, all with the same results.
Is there some good documentation I can read up on to figure out what I'm doing wrong? Thanks. (-:
P.S.: So far I'm very happy with the new firmware!
Edit: I didn't accurately describe what was happening. AC PtMP Mixed mode and AC PtMP mode with flexible frame length were both performing about the same, and both were much higher than with version 8.0.2. Setting a fixed frame length with or without asymmetrical duty resulted in dramatic (about 80%) reduction in capacity.
a week ago
The GPS Sync Design Guide has some good basic information, but I'm not sure we understand all the variables you are working with based on your description above.
We have been told that Mixed Mode currently uses the latest and best flexible frame code, whlie Pure AC mode currently lags a bit.
Fixed frame mode will deliver less download compared with a lightly loaded flexible frame AP, but latency will be much more predictable, especially as the load increases.
Is it really mixed mode that does it? Isn't it PtP vs PtMP, AC PtP is better then AC PtMP?
At least in my own basic test I get this result if I remembered right and AC PtMP vs mixed mode PtMP is not so much difference.
I also read somewhere that it might be a interference problem and that PtMP is better at handling it then PtP. Don't know if that is true and that PtP might still be better is a really quiet environment?
@UBNT-James This kind of stuff seems so "well known" on this forum (for a small group of people) but for new ubnt users it is very hard to find information. It would be nice to have a sticky thread for new users with all known quirks, stuff comfirmed/under fixing process (no need to report) and other gotchas/good to know.
Oops, I didn't properly describe what I was seeing.
AC PtMP mode with Flexible frame duration appears to work pretty much exactly as well as AC PtMP Mixed mode. Setting either a 5ms or 8ms frame length dramatically reduces capacity, even when I try different DL/UL ratios.
We're noticing a few similar things too. We've been running the RC client for a while and had fixed frame w/ sync turned on for on AP and rolled it out to two more, with about 40 clients between the two AP's last night. This screenshot from aircontrol sums it up. Left of the capacity charter is flexible frame, rightof the chart is after turning on 5ms fixed frame at 67/33. I was pretty disappointed to see the capacity get cut by 2 to 3 times, but since this is realtively new, I don't think anyone knows the real ins and outs yet.
ubnt, any insight on what we should expect in this type of situation?
Could just be the change in capacity reporting:
- Stated capacity in fixed frame mode is the sum of RX and TX, so the max is really the channel capacity.
- Stated capacity in flexible frame mode is the average of RX and TX. Since each direction can use all available bandwidth momentarily, it looks higher.
@UBNT-SNK I didn't do any TCP tests as a client on the tower, but the built-in speed tests showed speeds in the single digits on links that were previously testing 40+.
I don't think I can afford to try messing with it right now, but I've got another AP that I'll be able to try it out on later.
@benvanderjagt Thanx. We will wait for your comment on the other AP when you try it. In the meantime if you want us to take a look at this AP where you reported the problem, I will be happy to. Just email me
[sriram at ubnt dot com] with the information.