Well, first understand bandwidth utilization is always either zero or 100%. I.e. either data is being transmitted or not. When we "measure" bandwidth utilization, it's a measure of the ratio between zero and 100% usage over some time period. So normally to measure bandwidth utilization you take a snapshot of bits/bytes sent at some time, another at a second time. The delta provide bits/bytes transmitted during that time which you can they compare to what could have been 100% utilization all the time.
For example, if we measure 5 million bits set during one second, and "know" it's a 10 Mbps path, utilization during that one second was 50%. (Also understand, this means circuit was 100% half the time measured, but we don't know how the transmission was done. I.e. perhaps link was busy all during the first half second, or only all during the second half of the second, or all busy the first quarter of the second and the last quarter of the second, or . . . All, would be calculated at 50% utilization for the measured second.)
For an end-to-end measure, we can do the an end point's stats too, but the utilization percentage can only be calculated if we "know" the maximum potential bandwidth end-to-end. If we don't know the maximum possible, we might measure forcing 100%. This requires we load the link such that we believe or know its at 100%. "Believing" maximum bandwidth is by sending a bulk set of data, injected at one end measuring how long it takes to be transmitted and received. Unfortunately, of course, we cannot determine potential bandwidth beyond the path's bottleneck (nor can we use it anyway) and it's possible other traffic, at any one time, will "hide" potential bandwidth. I.e. we can only determine the maximum available bandwidth at any one point at time. (Peak bandwidths, that all hit the same measure might reveal the maximum available bandwidth, but might too.)
So again, to "calculate" bandwidth we measure transmission/receive rates for some time period. To measure bandwidth utilization we need to also "know" maximum possible, and then the forgoing measure can be calculated as a percentage. Again, if we measure 5 million bits over one second, or 15 million over threes seconds, or . . ., we have a 5 Mbps transmission rate. If we know it's a 10 Mbps path, we also can say we've had 50% utilization.
Things like "ping" or SLA, or ways we have to generate artificial traffic we have something to measure when we otherwise would not, and something that transmits without some "unknown" application delay.
