Question : Discovering bottlenecks

We are currently trying to find out why we are having network congestion over our 1GB fiber link between two switches.  I'm currently running some counter logs on quite a few workstations and a server we feel the majority of the traffic is going to.  What are the acceptable figures for things such as bytes/sec, packets sent/sec, and output queue length?  Is there anything else I should be monitoring?

Why does MS scale each of these counters to different values.  For instance, some are at .1 and some at .0001.  If I were to truly measure a 1GB link, should I change the vertical scale of the graph?

Answer : Discovering bottlenecks

www.neon.com have a free (trial one week as I recall) analyser called cyberguage that monitors graphically two end points. Also ethereal (www.etheral.com) plus other packages such as Solarwinds, ptrg that will be better suited to track network congestion rather than the performance counters on the servers. (My opinion only)

personally i see anything that hits a 60% marker as a bottle neck but it depends (all are different) on your traffic model. What to me would seem horrendous might be perfectly correct for you.

use the tools mentioned above to get a baseline to work from. These will also show you the main 'hitters' and then its case of logic/common sense to isolate out the one(s) that are causing you a problem in respect of bandwidth, queueing etc.