This is the minimum one-way or round-trip latency observed across the sampled test frames at this frame size, giving the best-case delay figure for the sub-test. An admin compares this to the maximum and average to judge how much the delay varies, since a minimum latency close to theoretical wire-speed propagation delay but a much higher maximum would point to intermittent queuing delay rather than a fixed processing overhead. For example, cienaCesBenchmarkGenTestSessionLatencyResultsMin reading 12 (microseconds, per the object's Unsigned32 type) on a local switch hop represents the essentially unavoidable serialization and switching delay for that frame size.