ONTAP Hardware

LUN ThinProvisioning performance



Just a general Q

For applications which we want to guarantee the best possible performance from LUNs - how much, if any, additional performance should we expect to get on a thick provisioned LUN vs a thin provisioned LUN ?

i.e. LUN space reserved vs LUN space not reserved





Another thread started a few days ago on this. None of us on the thread can see how performance would be affected either way.


suppose it comes down to the process that dataontap uses to decide where to write new data withn the volume (which i dont know in depth)

i assumed that if the LUN was Thin Prov that it may take a 'fraction' longer to allocate / decide on space and the write to the volume ?  (maybe nothing in this ??)

although - i presume that if a volume is containing only 1 LUN with vol Frac res set to 0 and that volume is guaranteed to volume then it makes no odds to space if the LUN Thin Provisioned anyway ? (as no other 'activity' would write to, and use the space for, the LUN anyway)


Besides lun container accounting in the volume, I don’t think there is any change in how writes are done guaranteed or not.


ok thanks

as soon as i said it the dba was all over it so it made the customer happy

as long as they have their data and like i said no other LUNs in the volume then i guess it's not hurting

LUN reservation will be useful if not doing vol guarantee (so giving free space to the aggregate) or on rare occasion if you have multiple LUNs in a volume ?


Good point…where to do the thin provisioning… sometimes to the aggr which means no guarantee on the lun or volume in that case to give the space savings back at that container.


Using thick vs. thin provisioning on NetApp is IMHO a question of personal preference.

Thin provisioning on all layers ramps up the efficiency (only actually used space is taken form aggregate), but requires solid monitoring tools (or checking aggregate free space on a regular basis )

Just a side note:

Some other vendors do have performance penalty when using thin provisioning - e.g. EMC VNX




correct, ONTAP has no performance impact when thin provisioning - since you don't change the way data is written!

With most other vendors, the fundamental way of writing to disk changes between thick vs thin.