Software Release R22 is now available and is the latest version of software for the PMP 450 platform. It contains a feature called Multiple Groups per Frame or MGPF. You can read about what it is and how it works here:
I would love for folks to download and update their networks, and gather some feedback about how it’s working for you, especially if you’re seeing any capacity gains and/or reduction in latency while using MGPF. It’s especially useful on busy sectors and during heavy traffic periods. Please post your responses in this thread. We have already heard from a few folks that were testing this while in Beta, but now that it’s out officially, we’d love to hear more!
We updated all of our 3.x 450M APs to 22.0, and utilizing 5 MS Frame Period with a 80/20 Frame Configuration, we have seen increased throughput and a decrease in Frame Utilization. On one of our 3.x APs using Link Test with Multiple LUIDs, total aggregate throughput is now over 500 Mbps. This AP surpassed 400 Mbps download last night, although the MU-MIMO Frame Utilization flatlined. Does the MGPF use the CPU? We want to make sure that we are keeping an eye on the correct stat for the other APs we have in the system. But overall this fix helped a lot of our APs that had the frame rate hitting 100% and flat topping.
If you’re using cnMaestro you should be able to track before/after CPU usage on the Performance tab. That being said, we’re using R22 on 450m 3GHz AP with over 100 SM’s registered and a 40MHz channel and CPU usage is just under 75% during peak hours, hitting around 260mbps aggregate.
We are using cnMaestro to keep track of our system and we saw a small jump in the CPU on our aps. The AP that I mentioned, I forgot to add that is was at 40 Mhz channel with a 5ms Frame period. How is the grouping on your AP? We saw our grouping change from a good solid 3 to a little more over the place
We don’t have many heavily loaded 450m’s. We have a 3 GHz 450m with approx 55 SMs. Prior to the upgrade it would max out around 200-225Mbps in the evenings. After, it is maxing out between 250 and 300M.
Here are downstream throughput and total user average latency statistics (measured from our data center to the customer end-devices) for one of our APs. You can see the before and after peaks.
We still hit full utilization and had latency increase, but there was noticeable improvement. Additional AP level shaping a little below peak capacity allowed us to iron out remaining latency during peak usage.
Most interesting is your graphs seem to indicate an actual congestion on the previous days your new peaks seam to indicate that’s not actually the peak as theirs no hours stuck at a capacity so even more capacity to exploit likely
