As the title says, I have done Link Capacity tests on several customers (various locations) with "Bridging and MIR" and receive ~5Mbps down and ~1Mbps up. However when the customer does a speedtest, they get ~1Mbps down and ~.5Mbps up.
What could be the cause of this? The APs (Cyclone Omni using Canopy 2.4GHz P9 AP) are fed by 100Mbps backhaul links and aren't overloaded (~30 subscribers on each AP).
Try setting the customer radio from 1x/2x to 1x (no rate adapt) in the General tab in Configuration. I have noticed that usually helps with better latency and throughput.
First, I would consider 30 clients on Canopy to be overloaded, unless they're all expecting ~1Mbps. During peak usage (evening video streaming and gaming) they could certainly be trying to pull 2Mbps+ each for a 60Mbps+ load per sector. (capacity oversell ain't what it used to be) Not saying that's your problem, however.
Two factors - "Bridging and MIR" test basically reports the max raw data rate it can achieve, not allowing for sector load or pesky 'real-world' things like TCP handshakes. And you don't say where/how they are running a speedtest.
On our network (YMMV) we've found the best results with http://speed.googlefiber.net. Customers provisioned for 10Mbps (ePMP) who could only get 4-6Mbps on speedtest.net would see 9-10Mbps on the Google test. When you click to run the test, it will warn that you're not actually on Google Fiber, but the test will proceed. (we also run the speedtest.net mini on a server here in our NOC, but discourage customers from using it since the server hosting it doesn't really have the horsepower to handle several simultaneous tests, to which they were sometimes subjecting it)
To be clear, we run mostly 900MHz Canopy, 2.4GHz/5GHz ePMP and 3.65GHz PMP320 (also some PMP450 3GHz and 900MHz, but too few to count) - while we have some 5.2GHz and 2.4GHz Canopy gear on-hand we pulled it all out of service several years ago.
j
@JLoewen1 wrote:Try setting the customer radio from 1x/2x to 1x (no rate adapt) in the General tab in Configuration. I have noticed that usually helps with better latency and throughput.
I would discourage this unless it's all that works - setting rate adapt to 1x means it's using the same timeslot at the AP for half as much throughput. If you have a power user set to 1x they can cripple a sector's capacity.
j
Thanks for the input folks.
For the sake of brevity I didn't include some things, including speedtest location; this is done using either our local speedtest server or via speedtest.net. Both have consistent results.
The reason I say it's not overloading the AP is because when looking at the SNMP graph on Cacti, it hasn't approached the AP capacity/plateued during peak hours. Also, this issue is present on an AP with less than 10 users and a 100Mbps backhaul link.
As for 1x/2x, many customer's aren't using Advantage SMs, so they are limited to 1x. I'm a bit confused about your statement "if you have a power user set to 1x they can cripple a sector's capacity". I'm a bit naiive when it comes to this setting, but from my understanding that's almost opposite of what you should expect from 1x, and that 1x/2x should only be used in certain/optimal conditions. Am I way off here? Why would this be?
2X, if workable, should permit twice the capacity of 1X within the same timeslot. So if a customer is pulling 2Mbps constantly, 2X would use half as much of the APs available timeslots as 1X would for the same data transferred. (it's more complicated than that of course but that's a good characterization of 1X vs 2X) If you are using all available timeslots then you've capped the AP, regardless of the bits/s (or packets/s) actually carried. (the only way to increase bps/pps at that point would be to go from 1X to 2X, or more generally to achieve a higher modulation scheme, as 1X/2X/4X/6X/8X represent)
j
Sure, but the negative effects of enabling 1x/2x outweigh any perceived benefit and make it unpractical for standard deployment. At least with our current equipment.
Putting aside 1x/2x, I still can't find a reason why the speeds would be so different with so much available bandwidth; especially in non-peak hours.
Hi Newkirk;
What is your experience and setup like for the Canopy PMP100 900Mhz series. Also the ePMP 2.4GHz series. I guess I should be asking about the ePMP experience in the other forum area.
For the Canopy 900 are you acheiving like 1.6Mbps on average. And say distances like 1-3mi, 3-5mi. RSSI is like? We are using 2 x horizontal omnis. One by Pac Wireless and the other by Last Mile Gear. Same tower. Use the two really for redundancy. We found relying on one was not good.
Regards.