I'm curious as to how the APs behave when saturated. If an AP is saturated, and all the SMs on it are trying to pull traffic, how does it distribute between them? Does it give out an equal amount to each SM? Does it scale how much each SM gets according to it's QOS values? Does it take into account what modulation an SM is using? Is there a way to increase an SM's priority so that it receives more bandwidth than other SMs when the AP is saturating?
I've got a customer who wants faster service, and I want to know what the implications of changing her settings would be in a worst case scenario.
When we start to see high frame utilization on the downlink and/or uplink, we start seeing latency increase across the board for all SM's. SM's with poor modulation will start to see more jitter as well. I believe the way Cambium's TDD works is that it gives all clients by default an equal time slot, and then on SM's with poor RF, it tries to retransmit using ARQ until a failure occures which results in a dropped packet. This ARQ retry period typically does not effect other SM's.
In the QoS on an SM, you can enable "Hi Priority Channel" and set an downlink/uplink CIR. Doing this actually allocates another control slot to the SM, and gives BW priority to that SM at the expense of BW from vanilla SM's during periods of AP saturation.
Interesting. What if I was to separate the two, and either assign a high priority VC without setting a CIR, or setting a low priority CIR without assigning a high priority VC?
Hmm, I actually have no idea if home networks use diffserv at all outside of VOIP applications. Do individual computers or consumer level routers actually assign code points based on ports and such, or do QOS settings on routers only apply for how the router distributes it's own traffic?
The customer has 10 people in their house, so I'm thinking QOS is something they need to set up on their home router anyway, since the real issue is the kids hogging the connection and interfering with the mother working from home.
Of course the kids would surely say that Netflix and BitTorrent and Warcraft and what-not ARE priority... :)
However, the bigger picture would be that setting QoS on the customer router would ONLY AFFECT UPLINK. Well, downlink from the router to the customer device, over LAN or WLAN, but it's likely that the immediate cause is on the WAN side. (QoS on that router can only truly control packets outbound from the router -downlink traffic from the world to that router will have already passed through AP and SM before reaching the limits/controls enabled on that router)
If the customer (or the sector) is suffering from uplink saturation then that can help, but on our network downlink saturation is almost always reached first. (whether saturation of overall sector capacity, or of the provisioned service to a particular customer)
It would be more useful overall to apply QoS at your border, inbound, tagging obvious VOIP highest, obvious fileshare lowest, obvious video stream low/lowest, etc.
It's a complicated prospect trying to prioritize DOWNLINK traffic based on which user within the LAN is receiving it. And prioritizing UPLINK, while reasonably easy at the LAN/WAN demarkation, is probably NOT by itself going to help.
Separately, at the LAN/WLAN level if there are literally 10 people, I'm guessing there's at least 20-30 devices on the network between PCs, game devices, phones/tablets, printers, etc etc. Which could easily lead to congestion on wifi if we're talking about a single retail-grade router. (I'm surely an outlier, but at home with just myself and my son we have over 20 devices on wifi, currently spread across a pair of Unifi APs)