Our favorite trace interval is 2.5 seconds. This will catch most problems with a minimal impact on network bandwidth. There are a few situations that may call for more frequent samples, but for a general setting (VoIP and otherwise), 2.5 seconds is perfect.
As far as "Samples to include", this depends on what you're looking at. Samples to include does not affect data collection - it's just a reporting parameter, so if you normally set it to something low (10, for example), but then later want to look at a slightly longer period, you can do so. Keeping it low most of the time is good because it's easier for the calculation engine to do it with fewer samples.
Once you're seeing problems, then you can modify Samples to Include to try and capture the entire problem period. You can do this by double-clicking on the end of the problem in the lower time-graph to focus the upper graph, and then entering a time-period in "Samples to include" (like "1 hour"). The display will then figure out how many samples to show and convert for you.
Hopefully, that points you in the right direction.
- Pete
