This is the average per-interval multicast-packet delta over the sampled window for this interface. A rising trend in this average over successive polls, even without any single alarming spike, is the pattern an admin would use to catch a slowly growing multicast workload (like an expanding cluster) before it becomes a bandwidth problem. A deltaAvg that doubled over a week on this interface would coincide with, for example, additional nodes being added to a multicast-heavy cluster application on that server.