I've been going through again the handy "updated library of custom vuC functions" authored by Scott Barber and three other well-known gentlemen.

I have to test in a setup where their ConnectSpeed routine will be useful.
I'm wondering though how their promised "linespeed" function would differ from the "connectspeed" one.

"Connectspeed" sets the line_speed at the beginning of a series of steps.
If multiple vusers perform the same series of steps, wouldn't each of the vusers connect at some randomized speed?
If so, how would "linespeed" be different?

The setup I'll use the routine in(I mention this in case somebody has a different suggestion or advice) is so:

Application/web portal/servers/DB stored in a datacenter.
Internal users access the application and the web portal by going through a VPN at 12Mbps. The VPN connects to the datacenter at 8Mbps.
In addition, the datacenter allows access to the web portal from the internet. This connection is at 6Mbps. Internet users of course use a variety of speeds.

The bulk of the testing will be on-site at the datacenter.
Besides the standard tests at normal LAN speeds, I'll need to simulate a number of internal users at current real-life speeds (max line speed of 8Mbps since I don't think the extra 4Mpbs on the user<->VPN link will be relevant).
Similarly, I'll test for a number of external users at various speeds including their current real-life one of 6Mbps.

I'd like to run some tests too by setting up some agent machines (or the control machine) on the User's area and have it access the datacenter through the VPN.
Would TestManager/Agents have a problem with that? I haven't found many references to VPN concerning Robot and I'm not comfortable with that.

Thanks for any advice.