Looking at the example given by HFSS manual, I can see a ring around the inductor acting as a gnd reference for the port, but it is not connected to the actual gnd underneath(below the silicon), so how does it work? Should the ring be connected to the gnd below in the actual implementation?
The ring represents a closed metal boundary in the chip, where return current can flow in real hardware. So that gives you the dimensions: same as hardware. It is important to understand that this ring forms a shorted secondary loop around the inductors, so it has some negative effect. Unfortunately in HFSS you can't model inductors without such a return path, and on chip (hardware) you also need some return path.
The trick is to have "enough" distance between the inductor and the ground frame, and "enough" depends on multiple factors like number of turns etc. Note that larger distance from inductor to ground ring also means longer feed lines, which are included in results and degrade the Q factor.
Sometimes designers use a slotted frame (not closed loop) when inductors have both terminals on one side.
This inductor designs is somewhat tricky, and it can take time to get a feeling for "good" methodology.
If you want to measure the resistance of the full path (including return path in ground frame), then yes.
Obviously the width of your ground frame matters in this case.