Hi experts,
A customer has run a test with 64 byte packets and experience a 50% loss. This was on a PMP450b connect to a PMP450i AP. It seems that the terminal's CPU does not support constant 64 byte traffic, and in a point dies....
Is this normal behaviour?
Hi Sofia,
The PMP450b radio has 50k pps processing power (this spec is in the PMP450b spec sheet). This can be a bottleneck if the packet sizes are small and the throughput is high. For example, if the modulation is high and the channel size is 20Mhz then the PMP450b is likley to hit around 120Mbps. To determine the smallest packet size that can be transported before the CPU PPS becomes a bottleneck you have to take the link speed and divide it by the pps to determine that packet size. 120x10^6 bits/sec / 50x10^3 packets/sec / 8 bytes/bit = 300 bytes/packet aka 300 byte frames. So, yes, if you're running a link at full speed then 64 byte frames will results in packet loss because the CPU will not be able to keep up with the small packets coming through the link.
One thing to note, however, is that typical internet traffic is about 576 bytes on average, so the pps is sufficient for most internet applications even for links running at full speed. If you have applications that you know are comprised of small packets and that small packets will comprise a majority of the traffic being transported on your PMP450 network then you can look at SM's with higher pps power like the PMP450i.
Thanks for your answer!
In terms of CPU, does PTP450i has the same characteristics as PMP450i?
We don't have PMP450i SM, but we can manage to change it for a PTP450i if this solves the problem.
Our customer requires the equipment to pass this 64byte test, but I agree that this is not a very typical scenario...
For the 5ghz and 3.xghz bands, yes the PTP450i has the same if not better PPS then the PMP450i (due to PtMP overhead)... BUT you cannot use a PTP450i as a client for a PMP450i AP. You will need to buy a PMP450i SM.
If your client is very serious about this 64 byte test, you might want to also consider selling them a PTP670 link which has the best PPS of all of these radios with a line rate of 850k PPS vs. PTP450i's 50k-ish? PPS. (Cambium staff please correct me if I'm wrong, I'm basing this on older comments and spec sheets)
Yes of course I know that we can't use a PTP450i with an AP... The idea was to change client equipment and install a dedicated PTP link for this specific client.
We dont have PTP670 to offer, nor PMP450i. The workaround I find here is the PTP450i that we do have in stock.
So, is it guaratee PTP450i supports 64-byte constant throughput? Or shall I test it first in lab before offering this as solution for our client?
@Sofia Perez wrote:
So, is it guaratee PTP450i supports 64-byte constant throughput? Or shall I test it first in lab before offering this as solution for our client?
I'd try first in the lab if I were you.
I agree with performing a lab test to confirm this. I recall seeing 75k pps as a target pps processing power for the 450i, but I'm not sure if we've achieved that yet. Even then, it's only 50% faster than the 50k pps of PMP450b today so it won't be able to pass an RFC2544 test with 64 byte frames at max throughput in a 20MHz channel. What throughput are you testing up to? If you're testing higher speeds I'm afraid the PTP670 or PTP550 will be the only platforms that can pass the RFC2544 test with 64 byte frames at higher throughputs.
The pps you require for 120Mbps speed with 64 byte frames/packets is 120x10^6 bits/sec / (64 bytes/packet x 8 bits/byte) = 234,375 packets/sec. The PTP50 has 250k pps and PTP670 has nearly 1M pps so those would be the only platforms that could manage the processing load.
For PMP450b we are testing 10/10 Mbps throughput (20 Mbps aggregate) and throughput is limited to 5Mbps for 64byte frames.
If we do the math, 64 bytes shall pass with aprox. 25Mbps throughput without reaching the CPU limit, am I wrong?
I've attached some images of the test ran with Link Test tool.
64 bytes --> 9.81 Mbps (aggregate)
128 bytes --> 20.63 Mbps (aggregate)
256 bytes --> 39.83 Mbps (aggregate)
Also, our client tested RFC2544 with 50% loss for 64bytes, which is coherent with the 5Mbps in Link Test results.
We are going to set up a lab test in order to test PTP450i behaivour.
Hi Sofea,
Do you have the testing result for PTP450i?
