We have excelen test site. Sector A has 15 subscribers, sector D has 45, so totally 70. All of them are int the area about 120 degrees.
I droped users from havy loaded sector, than start to move users by 5 back. I did measurement by wireless link test from one user. There are link speed degradation correlated on number of users.
According to the monitoring total upload utilisation never rich 5,9 Mbit which is much less of capacity of the system.
whats your channels width, number of subs and your %%% break down per modulation. also from your CPE side, what is your link quality and capacity on a few of the weaker subs? the excel sheet for the bandwidth calculator does a decent job forecasting bandwidth. also as you add subscribers, if they are all busy they will split air time evenly so you will see a drop in speed with a number of subs, if you've got a lot of extra downlink frame, or your DL frame usages is low, you can correct your uplink bottle necks with that. also if you've got erratic pings, hitting one sub at 30 ms for 5 or 6 in a row and the next hits at 90 to 140, that packet was likely retried, noise may be a factor in your performance issues. check your performance tab for a break down of TX and re transmitted packets. theyre was some under the hood improvements in 2.4.3 so i'd suggest upgrading your firmware to the current 2.5.1 and you'll see a performance improvement. the uplink ussage will still falsely read busy FYI
Any idea what the upload traffic is? On my residential network I'm seeing about 5-6% of my traffic as upload, aobut 1/3 of what you're seeing. If, for example, the upload data was a bunch of small packets, it could saturate the radio without seeing the kind of throughput you'd expect to see. Two things that come to mind are bit-torrent users and various viruses/bots trying to do DoS (DNS, NTP, etc).
From you ap side, what are you uplink mcs %% per code state ? For example, mcs15 - 15%. Mcs14- 20% mcs13 - 20%… If your average code state through put is mcs 12, your on target, 13 your close 14 would higher yet. Use the capacity planning guide to check your results. Our links are very close to expected. You’ve also got to take into account your re-trasmits ect. 10 mb depending on your average code state, and error rate isn’t exactly bad for a fix frame ratio on a 20 mhz wide signal. Mcs15 with near 0 error rate will land you a little over 20 Mbps
our highest connection count is 58 at the moment, its in 40 mhz mode though. we've got a much wider MCS mix
we've got customers connected at MCS 12 and getting 11 to 13 mbps from it when the sites quite, durring peak its 8 to 10 on the speedtest.net (sometimes less when load is very high
how are you measuring your speeds ? I'd suggest rather than using the wireless link test to judge whats happening, use your SNMP software and cause actual traffic you can watch, send a large file up. if you've got background traffic, I don't believe the link test will kick aside the other traffic. you may be hunting for a problem when there isn't one or isn't much of one.
check your uplink re-transmits too. if you're getting any big % of those you're going to start loosing uplink. I know our uplink on the attached sector isn't doing hot as it clearly shows and we are getting as expected performance considering link conditions.
