HPROF analysis with MAT in the context of CONTROLLER-1756 is showing 4 MB used up in org.opendaylight.genius.interfacemanager.InterfacemgrProvider, at a stage when (according to Sridhar/Sai) the system should be "at rest" - so this looks suspicious.. so I'm guessing something is probably not really done right there.

Now 4 MB is not huge in the large scale of things, so this isn't a Blocker, but it still doesn't look quite right that in a "Dominator Tree" in MAT after a lot of cluster related shit this class (only, nothing else in genius or anywhere in netvirt) does visibly stick out as consuming a surprising amount of memory - it should, ideally, not hold on to permanent state (whatever it is) like this?

In theory, I suspect that this would over time, gradually and slowly, probably cause OOMs. Not sure under what exact conditions (lots and lots of interfaces?), but if we have some unbound Map or whatever somewhere in there, it would be good to have a look at it (PERHAPS using a real [infrautils] Cache, instead of a Map; if that's it).