Hi
Charles,
Your
DBA doesn't understand what qtrees are about. Not his fault, he's probably used
to working with block I/O storage.
A tree
quota applied to a qtree will incur the same performance hit regardless of the
state of the qtree "capacity". i.e. every write will require a hash table
comparison, which may cause it to respond with a "disk full" if the quota limit
is reached, or an alteration (usually an increment) to the hash
table if the write is completed successfully. One of the NetApp guys can
provide a more accurate technical description, but that's the general idea. You
can resize the tree quota both ways in front of the DBA to help him understand
that tree quotas have nothing to do with the layout of data on the disks. (i.e.
increase the tree quota so the qtree is only 40% full, run a DB test, shrink the
tree quota so it's 99% full then run the same test).
However, he is correct of course that there will be a
performance impact when the parent VOLUME is nearly full.
Hi,
We're running a test oracle DB connected to a netapp
filer.
Our DBA is convinced that when the qtrees are close to full
(>90%) the performance goes down despite the fact that the parent volume is
only 50% full.
Does this make sense or is it more likely the apparently
slowdown is coming from somewhere else?
-C.