Automation of traffic latency measurements

Hey everyone,

I help manage a few small fiber DIA providers (we're talking a few DWDM muxes with fiber out to clients with a small metro switch at the end point) and am constantly trying to dial in the monitoring systems to be more useful than noise and one of my biggest gripes is "How do I reliably measure throughput latency on the last mile?" Now, many others just refer to measuring ICMP responses between core and edge but that would be a terrible metric since it's all processed via CPU and not necessarily reflective of actual traffic being processed. So my question is; how does everyone else measure the throughput health of their last miles for SLA purposes? Y.1731 maybe?

Thanks