yeah you're gonna have to define low latency here

what are you doing differently to solve the predictable latency problem? having all the processing happening in the compute cluster so that there isn't enough back and forth to the ground to notice?