Start a Conversation

This post is more than 5 years old


Go to Solution


February 10th, 2015 16:00

VNX Performance Thresholds


I am currently working on a hospital project, we have three storage pools, of which 2 are over the 80% consumed with the third at 65% but oversubscribed an additional 50%.

General rule of thumb has been on most vendor SANs that performance thresholds are around the 70-80% capacity for optimum performance before seeing degradation of performance moving forward from there.

I have not been able to find anything in any technical or white paper with EMC on were the best performance thresholds are, only that the SAN alerts at 70% and one of these discussion threads showing that a second alert at 85%.  I have a very slow procurement system, I am currently identifying with my management that this is a risk if we continue to grow on the final storage pool - they want a formal answer on where degraded performance can be seen.

Can any help me with some advice where I can find any reference to performance thresholds.



4K Posts

February 11th, 2015 19:00

When a Virtual Provisioning Pool runs out of space, new slices cannot be allocated and host writes cannot written to the Pool LUN.

FAST Virtual Provisioning requires 10% of the slices to be free in Release 30 and 31, on the drives/tier it would use to relocate the data to.  If there is not sufficient free space on that tier, no relocations can be done into that tier; however, if other tiers have sufficient free space, then relocations away from the tier with minimal free space can still succeed.

The performance of a Pool can also decrease, if the relocations cannot be done, because the busy slices would build up on the high capacity drives, rather than being moved to high performance drives.

In Release 32, the relocations can complete when there is under 10% free space, but not having enough free slices, would limit the amount of relocations that can run at once.  Therefore, for Release 32, best practice is to retain at least 5% free space on the Pool.

For more details refer to "VNX: Storage Pool does not have enough free space for FAST VP to relocate slices."

1.3K Posts

December 8th, 2017 04:00

Capturing the performance difference would not be an easy task. But after 97% you would see drastic difference in the response and throughput, disk rebuild time. The behavior is slight different towards a random versus sequential workload. But keeping below 95% would be always better.

No Events found!