[rrd-users] Sanity checks on Buffer Cache and I/O at peak aggregation times
rick.jones2 at hp.com
Tue Sep 27 20:58:19 CEST 2011
> the 'problem' is that a disk block which is only hit once a day
> will most likely not stay in buffer cache, so as rrdtool goes in to
> update the slow RRAs, it will first have to read the block in order
> to then update it ...
It may not be sufficient, but under Linux at least there is the
posix_fadvise() call - might that enable rrdtool to request the cache to
retain file sections of particular interest?
More information about the rrd-users