How is latency measured?

Hi there, curious developer here.

I was reading an article on Tech Radar about the Grid. It’s from 2013 but I would imagine it’s still relevant today.
http://www.techradar.com/news/computing/servers/how-nvidia-grid-is-set-to-revolutionise-cloud-gaming-1167986

In particular, I’m curious about this graph:
http://cdn1.mos.techradar.futurecdn.net//art/internet/Nvidia/Grid/nvidiagrid%20(6)-650-80.jpg

How exactly was lag being measured? I see the Grid solution had a latency of 161 ms while the GeForce PC had one of just 65 ms. Are you measuring how long it takes each frame to complete from input device, through the update/render pipeline and to display?

Cheers