Remote Desktop software: Why is Teamviewer so much faster than the open source competitors?

Question

TeamViewer is faster compared to FLOSS alternatives concerning 2 aspects:

1. The whole configuration process before the actual session: Determine IP, traverse through NAT & firewall, etc whereas in TeamViewer simply entering an ID/PW, which goes through the centralized system of TeamViewer (that's of course an issue of trust!).
2. The remote viewing/control session itself: VNC solutions are far less fluent and responsive, often lag back dozens of seconds whereas in Teamviewer it feels almost "live" except of course if much screen content changes at high frequencies (animated contents, etc).

I would appreciate technical explanations concerning #2. I can't just figure out where's the optimization potential, which makes TeamViewer so much faster.

Answer 1

VNC is just inefficient - it uses RFB. which in effect takes snapshots of your system every so often, and tells the client system to draw boxes of X size every so often (a full description of the RFB protocol is here. More efficient protocols don't work at frame buffer level - RDP uses a kernel level driver in windows, for example and nx caches and compresses data.

More efficient protocols would likely as such

1. cache data and only send the differences
2. hook into the native subsystems of the host os
3. using caching and compression to speed up data transfer.

In the case of teamviewer its likely clever use of the first and third - since it seems to use the same client for windows and linux, and uses their server to handle connection and routing of traffic