If the shape of the distribution looks like a bell curve or truncated bell curve, then the distribution is approximately log-normal, suggesting that the OTUs are participating in an interacting ecosystem. Other mathematical distributions can be recognized if the shape is different (Edgar and Flyvbjerg, manuscript in preparation).
Setting an OTU abundance cutoff
Octave plots can indicate a good choice of OTU abundance cutoff.
read depth (sampling effort)
Octave plots can indicate the number of additional reads needed to capture missing OTUs.
If the distribution appears to be truncated, e.g. if part of the left side of the bell curve is missing, his indicates that the full diversity has not been sampled. See log-normal distribution for an example.
If a clear peak is visible, then on the assumption that the distribution is approximately symmetrical, this suggests that more than half of the diversity has been sampled and it is reasonable to extrapolate richness using an estimator such as Chao1, mirror or FE. If the peak is close to the y axis or absent, then it is probably not justified to extrapolate because there is not enough data.