What with all the recent hubbub about Google taking site load times into consideration and all, I've been using Pingdom to analyze my site and need some help understanding the results.

Example link: http://tools.pingdom.com/fpt/?url=http://neosmart.net/forums//&id=2230361

According to them, the portion of the bar that is yellow is the time between "start" and "connect" and the portion of the bar that is green is the time between "connect" and "first byte" with the blue section being the actual transfer time (time between "first byte" and "last byte").

Unfortunately, they don't seem to explain what actually happens in each stage, so I have no idea what I'm supposed to be trying to optimize. My working assumption is that the first section of the bar is "network-level" time (time for the packet to reach an open port??), the second section is the "software-layer" delay (for the request to be processed by the webserver), and the 3rd part is pure transfer time?

I need to know if changing anything software-side (w/ regards to the webserver stack itself) will improve the first section of the bar, or if that delay is purely due to the overhead of running the webserver in a virtual machine.