How does being behind a Reverse Proxy affect FxM Access speed metrics?
It depends on how the reverse proxy is configured. If it is just a pass-through configuration where it is just rewriting the client IP before forwarding onto the webserver, then we should be able to detect the access speed. If it is a reverse proxy where it opens new connections on behalf of the users (and perhaps multiplexes multiple clients onto the same connection and even caches some results) then that is likely to affect the access speed calculation.
Access speed calculation is very sensitive to network delay: we basically take the bytes sent by the server and divide by ½ the network delay before we see an ACK for those bytes. So if that network delay is too short (ie ACK comes back too fast), then you are almost dividing by zero which causes the resulting value to be huge. Once the network delay gets too fast, we do not include those results in the access speed hence why it appears to be zero.
So, in other words, if the reverse proxy essentially acts like the "client" in the traffic we are monitoring we lose the information we need to calculate access speed. Consult the FxM Metric Reference Guide for an explanation of how access speed is calculated. In particular, we need the client acknowledgements (ACKs) in the TCP layer to measure access speed. With the reverse proxy configured as described the ACKs we see are coming from the proxy and not the real client. Hence, the information we need to calculate the metric is lost. You may see some users in the metrics that have a non-zero access speed but this is a reflection of the proxy-to-server traffic and not the client-to-server.
With regard to how other metrics are affected by this configuration:
If the proxy is opening connections on behalf of the client, then that is going to make the network delay calculations result in smaller values which in turn affects end to end time. The effect is likely to be insubstantial though (less than 1%).
This is a known limitation that cannot be corrected.