http://jira.hyperic.com/browse/HHQ-2589
While dealing with AvailabilityRLE issues we decided to chart the responsiveness (attached) over time of the various methods that we keep stats for. From the chart one can clearly see that every hour the process speeds up dramatically. At these times the only major operation is running and that is Postgres Vacuuming. Given the I/O (see attached) intensiveness of the operation and the speed benefit we should experiment with cutting down the interval to <= 30 mins and chart the effectiveness.
Anonymous