Measuring utilization
I have a cascaded switch network of 75 workstations where all the traffic is going to the Internet with a static NAT'd IP. I want to know how much Internet bandwidth is consumed by 50 systems.

What can I use (using MRTG) to see how much bandwidth is utilized at the serial port of the router?

    Requires Free Membership to View

    By submitting your registration information to SearchNetworking.com you agree to receive email communications from TechTarget and TechTarget partners. We encourage you to read our Privacy Policy which contains important disclosures about how we collect and use your registration and other information. If you reside outside of the United States, by submitting this registration information you consent to having your personal data transferred to and processed in the United States. Your use of SearchNetworking.com is governed by our Terms of Use. You may contact us at webmaster@TechTarget.com.

There are several means to measuring utilization of the link. The obvious one is the one you mentioned - sample the router counters on a regular basis with a utility such as MRTG. That provides a coarse-grained estimate of the amount of data passed per time unit of sampling. This should suffice for most planning exercises.

The alternative is to use an end-to-end performance measurement tool to sample the true available bandwidth at various times to see how the pipe is actually being filled. My recommendation would be appareNet of course but I'm biased. Combined with MRTG or other point source readings, this is your best source of information. If you can access counters on your switches then you are all set.

The benefit of the end-to-end approach is to let you see if the router, or some other point, is acting as the bandwidth bottleneck. If 50 workstations are already saturating the link, then the MRTG measurements associated with 75 workstations are not linearly related to the measurements for 50. This makes it difficult to determine how your provisioning should scale. If there is no apparent saturation at any point, then you can fairly safely assume that your network utilization is approximately linearly dependent on the number of workstation.

Since there you know the topology of your cascaded system, measuring end-to-end along different switch paths can help you determine if/where saturation is taking place in the switch network.

This was first published in April 2003