2011-07-29 02:29 AM
Assuming you had a new filer (running ONTAP 8) delivered with all identical disks, is it best to put them all into one aggregate? If not, what would be the reasons for using two or more?
2011-07-29 06:12 AM
There are aggregate size limitations, depending upon filer model and OS version. Having separate aggregates also ensures that you can separate workloads (volumes) that demand a lot of disk I/O. I suppose if your aggregate is mirrored you should consider that deltas are likely to be larger and rebuild times would be higher, otherwise there seems to be little perceived benefit in having multiple aggregates...
2011-07-29 07:03 AM
In addition to what the others already mentioned it's best practice to have a separate root aggregate, if you have enough disks available. Reason being, that there's a chance, although unlikely, that the aggregate containing the root vol has to be wafliron-ed. Filer would be unavailable until finished... So with a 'small' root aggr (aggr containing the root vol), this unlikely event would nonetheless be dealt with 'quickly'.
Also there could be different aggr options, like stuff you want to SyncMirror (and stuff you don't want to be mirrored (??...), or (more likely) SnapLock (not yet available on ONTAP 8, though).
Other than that, I'm all for aggrs as big as they can be, with RGs as big as they reasonably can be (balanced!! i.e. all RGs the same, if at all possible).
2011-07-29 02:31 PM
Further to the points already discussed - larger storage pools (aggregates) will improve the overall efficiency / utilization rate of your enironment rather than having islands of unused space that are locked in smaller aggregates; it gives you the opportunity to consolidate remianing free space in a much larger pool.
Another reason for segregating aggregates is to make capacity planning easier for replication. In complex environments where you may have a requirement to replicate only a portion of a storage system it may be easier to split your replicated and non-replicated volumes by aggregate. This is particularly useful when you are adding more drives on the source to add more drives in the equivalent aggregate on the destination system; and / or you have a set of complex fan-in or fan-out replication requirements.