Please find the answers inline below:
1. Should I try to match the QOS bandwidth limit on the AP's?
A1: I guess this is dependent on your environment and how the APs are performing with the current values you have configured. However it wouldn't be a bad idea to match the bandwidth limit of the APs if possible since you want them to perform at their full potential. However if you want to throttle the traffic coming from those APs then you can define the limits accordingly.
2. Will I get better processing time?
A2: I think this is related to the APs and this purely depends on how the APs are currently performing. If you do match with bandwidth limit of the APs and then it would perform better.
3. Overall,,, is using an exact buffera bad idea?
A3: When you configure transmit-rate 'exact' then basically you are hard setting the limit without any ability to burst higher if the resources are indeed available. So I would recommend using this with caution and per design needs.
Please Mark My Solution Accepted if it Helped, Kudos are Appreciated too!!!
The idea of configuring CoS is to prioritize specific traffic over other traffic during the time of congestion. Ideally a CoS solution/implementation will keep the same conditions for specific traffic across the whole network and in both ways (transmit and receive), this is why we say that we have to implement an End-to-End solution when we talk about CoS/QoS.
Being said this, I will say:
R/ Yes, in the APs you are prioritizing specific traffic over other types of traffic and this should be the same in the rest of your networking devices in order to garantee the same treatment to that traffic all along your network. The correct values to be set are a complete different story because only you can tell the needs your traffic has.
R/ It will depend of the values you set. The more priority/transmit-rate/buffer you give to specific traffic, the less chances it will have to affected by congestion in the network.
3. Overall,,, is using an exact buffer a bad idea?
R/ Im unsure about the configuration you have set so far, are you currently using "exact" option without configuring any values for transmit-rate or buffer-size?
By default only two queues are configured. Best-effort, which is mapped to queue 0, and Network-Control, which is mapped to queue 3 if I recall correctly. These queues are configured in the following way:
Queue 0 (BE): Transmit-rate and buffer: 95%
Queue 3 (NC): Transmit-rate and buffer 5%
Is this the way you want them to be configured? If so, configuring "exact" option will guarantee that they dont exceed those values hence protecting the resources of the other queues.
As a side note, it is very important to understand that CoS works by prioritizing the sending of specific traffic over an interface ( based on schedulers/scheduler-maps), and this is the only point where we can affect the traffic, when we sent it. On an ingress interface I can only classify traffic and I cannot prioritize which traffic will be processed first, however based on this classification I will eventually be able to prioritize the way the traffic is sent on the egress interface. Im bringing this to discussion because if we are trying to implement an End-to-End solution, we need to take in account the way the traffic flows to prioritize both ways of flowing:
Flow #1 (AP-to-Internet): in the ingress interface, ge-0/0/13, we use classifiers to put the traffic on specific forwarding-classes that will be mapped to egress queues on the egress interface, ge-0/0/0, and based on the configuration on those queues the traffic that was put in to the forwarding -classes will be affected when it is sent out of ge-0/0/0.
Flow #2 (Internet-to-AP): in the ingress interface, ge-0/0/0, we use classifiers to put the traffic on specific forwarding-classes that will be mapped to egress queues on the egress interface, ge-0/0/13, and based on the configuration on those queues the traffic that was put in to the forwarding -classes will be affected when it is sent out of ge-0/0/13.
Whenever you are checking outputs like "show interfaces extensive" and verifying the CoS statistics, you have to think that those rates are configured for traffic being sent out and that if you see traffic increasing on a specific forwarding-class/queue it was because it was classified in that way on the ingress interface. Having this mind-set will help you a lot with your CoS configurations.