Just as a matter of course, you should keep stats or be able to derive pretty much everything. Off the top of my head, the kind of stats you should be keeping are things like:
of msgs, broken down by type and sender of message
local reads, writes, remote reads/writes, both by replica
resends
session timeouts
empty heartbeats as opposed to heartbeats carrying write aggregations
sessions, avg len, avg quiescent period before tags released
avg #remote writes. Since you are in the simulator, you might also
be able to come up w/ a measure approximating of
writes-above-optimal, i.e. above when the optimal movement of an object should have occurred from one replica to another. This will be
very useful when evaluating heuristics and timeouts
Just as a matter of course, you should keep stats or be able to derive pretty much everything. Off the top of my head, the kind of stats you should be keeping are things like:
of msgs, broken down by type and sender of message
sessions, avg len, avg quiescent period before tags released
writes-above-optimal, i.e. above when the optimal movement of an object should have occurred from one replica to another. This will be
very useful when evaluating heuristics and timeouts