- ONTAP 9
- In an optimally balanced FlexGroup volume, all constituents have roughly the same amount of data and load, and the volume can maintain that state while using a high frequency of local placement for best performance.
- A workload with a good balance of folders and similarly sized files would be able to maintain local parent folder placement while also keeping a relatively even balance of capacity.
- A less optimal FlexGroup volume might have some constituents that hold more or less data than their peers, or that are receiving much more or much less traffic.
- Workloads that have only a few folders with many files per folder, or workloads with highly variant file sizes can experience data usage imbalances in a FlexGroup.
- Capacity balance, however, is not the most important function of a FlexGroup volume.
- Instead, a FlexGroup volume functions best when there is a mix of local placement for performance along with capacity and inode count balance.
- We don’t want to sacrifice performance for the sake of perfectly balanced capacities across member volumes.
- ONTAP constantly monitors the ongoing state of the member volumes and adjust placement decisions based on the current state of the FlexGroup volume.
- Prior ONTAP 9.8, if one member volume is a little out of balance from the others, then it’s likely that no adjustments will be made to ingest.
- But if that member volume starts to approach 90% capacity, or has a capacity discrepancy that exceeds 512GB, then ONTAP makes more aggressive placement choices for new data to correct the imbalance.
- This adjustment means more remote file placement to other member volumes, which can have a negative (but potentially unnoticeable―roughly 5% to 10%) effect on FlexGroup volume performance.
- ONTAP 9.8 changed the 90% threshold for data placement urgency (where ONTAP forces more data to other less full member volumes much more frequently when a member volume hits 90% capacity) to a higher value (99%) so that FlexGroup member volumes that are growing in size won’t adjust their data ingest rate by very much until they reach 99% capacity, eliminating the negative impact of remote file placement in the earlier versions.
Best Practice 2: Stop worrying about capacity imbalances.
- A FlexGroup with a capacity imbalance across member volumes is not in itself a problem and should not be treated as such.
- Instead, look at capacity imbalances as a potential cause if a FlexGroup is not performing as expected or if the capacity imbalances are so extreme that you’re running out of space on your cluster.
- Be sure to engage NetApp support if you feel that your FlexGroup volume capacity imbalance is the source of a performance issue.