Thought this was a great feature. Doing a 2 second link test causes all SM's on AP to loose connectivity for 2 seconds and really annoying to VOIP users.
I do tests like this in this order:
Extrapolated: 22/13mbps
wait ~20 seconds
RF Link: 64/13mbps
wait ~20 seconds
Now try Extrapolated again.
Extrapolated 66/13mbps
For next 30 minutes or so the extrapolated test will be close to what RF was. Wait a length of 'unknown' time and you must run another RF Link Test to get the Extrapolated accurate again. Anyway to fix this? I have updated AP and SM to "14.1.2 (Build 10)" hoping its fixed but its not. This is on PMP450i AP to standard SM and 450i SM. I think I see it on PMP450 in 3.6 as well.
Thanks for the post. Extrapolated Link Test was designed to help exactly in the scenario you described. But it does have a caveat in that if it is run soon after the session has been established, it may show a lower throughput than the radio can actually support. We need to update the documentation to explain this.
The problem comes in that when a radio starts session it starts at the lowest modulation, 1X. As traffic is passed successfully across the link, after several frames the radio decides to try the next modulation, 2X. And it repeats this process verifying every lower modulation before trying the higher one. For this reason, even an initial RF Link Test can show lower throughput if the session has been recently established. To guarantee to get the correct throughput after session is established, the first link test results can be discarded. We do not know of a way to fix this limitation currently.
If the session has been up and traffic passing, then Extrapolated Link Test should give fairly accurate to the RF Link Test.
Are you seeing the modulation go down without a session drop only to come back up after RF Link Test? If so, that is concerning and something we need to investigate more.
Thanks for trying the 14.1.2 beta, as we appreciate the feedback.
This would be an SM that has been connected for days although it has very little traffic. I frequently do this test on my house when no one is home. A PPPoE connection is present and numerous NATed devices in house are doing some very minimal chatting. Likely less then 1kbps. I run the extrapolated test and it tests 20mbps down about. Try it again and typically the same. I run RF test and its right at 60mbps. Run extrapolated afterwards and like you say it is then accurate.
I have a couple problems here.
One, I have corporate users that have VOIP services running on this AP, 2 second hiccups are not a very good option.
Also, when installers hook up a new custommer I demand a link test before and after install to determine link health. Usually the SM has nothing plugged into ethernet port at this time except power. Most link tests we try to keep at or not too much lower then 60mbps down and 20mbps up on 20mhz channel. If a user tests 15mbps down and 3mbps up I do not want them eating up air time and would rather pass on that install.
Is it possible to have a 10 second version of the extrapolated link test that will use the first 8 seconds to excersize the link a bit then the last couple seconds too do the actual test? All without disrupting other users on the AP?
I reported the same issue when it was first released in beta.
I'd be happier if I could adjust the length of the extrapolated test to make it more consistent. Two seconds isn't enough time for consistent full link capacity tests, and I can't see an extrapolated test possibly being more accurate than a full test.
Couldn't the SM's constantly look at all there traffic and what data rate and error rate they are doing and give you fairly accurate link quality/rate? I could perhaps clear these counters with SNMP then flood the modem with 1500 byte ping packets for 10 seconds then read the stats again with SNMP?
Or perhaps instead. Flood SM with 1500 byte ping packets for 5 seconds to wake up idle modem and let it figure out modulation rate. Clear stats with SNMP command to SM. Flood SM with 1500 byte ping packets for another 10 seconds. Then read the stats and figure out average link speed based on them?
I cannot tolerate 2 second AP outages for RF link tests although they are accurate. I also cannot tolerate installers installing 450 SM's that have less then 20mbps link test. Two of those running 10 mbps at peak time will kill an AP.
The PMP320 APs had an MCS table which tracked how many bytes of data had been sent using each modulation. It was fantastic for tracking down low modulations causing issues, as I could copy/paste it into a spreadsheet and figure out which CPEs were causing problems.