Improving high ping times off of tower

We have a tower site that has a 5.8 sm bringing service in and a 900 full ap that has 46 subscribers on it. and every one on that 900 ap has high ping times and complains of slow speeds. We also have 2.4 aps on there and the subscribers on there have no issues. So my question is since we have a lot of subscribers on the 900ap (on the SM’s) ----> what if we set the download and upload burst allocation down because we currently use its default setting (500000) for both the upload and download. the Broadcast/ Multicast Uplink Data Rate is set to 1.

#1 recommendation: split the users between two APs. (Whenever we start to approach 50 clients on an AP, we start looking for alternatives. Ideally I’d prefer to keep it under 30 clients per sector for the 900mhz gear.)

Question: What uplink/downlink ratio do you have set, what uplink/downlink provisioning, and how many control slots? (someone may correct me here, but I think it’s pretty important with that many clients on one AP that it has control slots set to perhaps 2 - that’s where we set all APs on our network)


Also, what’s the traffic load on the 5.8sm bringing in service to the AP’s? If that is approaching it’s capacity, customers would start to see high ping times and slow speeds.

smwtech wrote:
Also, what's the traffic load on the 5.8sm bringing in service to the AP's? If that is approaching it's capacity, customers would start to see high ping times and slow speeds.

But one would expect that to affect customers on the other APs as well. (which does bring up the question: Are the 5.8APs fed through the 5.8SM as well??)


Currently are ratio is set at 20000/20000 and are control slots are 0.
We run everything through a CMM.
Just to clarify we have several 2.4 ap’s and do not have an issue it seems more directed at the 900AP and we have changed out the AP.
All subscribers are bandwidth controlled for max of 1500/512.

First thing I’d do then is try setting the 900AP (and possibly any that can see this AP from a distance) to 2 control slots. If that doesn’t address the latency you can revert, but I’d keep it set even if you don’t see a significant change.


Definitely get those control slots up - aside from registration, Canopy needs to schedule in upload frames from subscribers. With 0 slots there’s not enough time for the AP to talk to the SMs to find out which need to transmit data. Control slot setting of 2 is recommended for the range of customers you’re running - 0 is only appropriate for up to 5, I believe.

If you’re running 1500/512 to your subs then try a burst allocation of 8000/4000. That’s about 1 megabyte download at full speed before subscribers get throttled down to 1.5 Mbps. Until NetFlix made a debut up here this ran just fine on our network. By using a larger burst value like 20,000 kilobits (2.5-ish megabytes) less and less customer traffic is getting throttled. The default values of 50,000 kb effectively bypass the sustained rate settings for the average customer as the time between bursts on the average webpage is enough to reset that bit bucket… so they never get throttled.

If you’re having problems with streaming (feedback from tech support, network traffic analysis, or even just basic graphs will help you determine this) then you can set a slightly higher burst allocation with a lower sustained rate. When people on 900 MHz decided that NetFlix was the way to go we had instant complaints from people reporting incessant buffering as well as regular users who couldn’t surf. The lower sustained rate not only clamped down how much NetFlix could move, but since it was more practical to actually achieve that lower speed constantly, there was less buffering and the video was smoother. Of course overall available bandwidth on the AP went up too so surfers were happy again.

Something to think about

We did increase our control slots and adjusted our downlink data percentage and over a 24hr period have noticed a big difference (For the good).
Netflix is also an issue for us we will try to adjust the burst on those that are having a problem.

to futher handle the netflix issue…

we use juniper for our routing
our J series routers can “write in” DSCP or ip TOS tags to packets from specific domains, IPs ect. ect. as well as sent a MIR for each data sessions to each local IP, with that said we can force netflix to a low quality mode and use the dscp tags to control flow in ournetwork so our VOIP customers don’t get garbage on there calls while the AP is busy and surfers always get a fare share of bandwidth…

its a lot of work but it can stop phone calls.

47 subs on a 900 is just begging for problems these days…

we don’t exceed 20 in most cases (depends on what activity the indivual ap is) some times we stop at 10 :frowning:

we have started using the 320 plateform to fix some of these issue… its got a few bugs, but with the right tweaks it performs very well. we have been able to switch most of our 900 subscribers that are locked at 70 db or better to wimax. the wimax isn’t cheep but you can put 5 to 10x the subs on a basestation without customers complaing…