Approving a 40 TB Thin Request in a Tight SRP
Capacity review for a PowerMax 2500: SRP_PROD has 200 TB usable, 175 TB subscribed, and 25 TB free. Thin volumes account for 150 TB allocated; snapshot usage is 30 TB; compression ratio is 1.6:1; 18 TB of thin volumes have never been written. A new analytics project requests 40 TB thin and 25,000 IOPS with a 70/30 read/write mix. Workload Planner shows the array's peak front-end 95k IOPS, back-end 70k, and front-end director peak utilization of 68%; the service level says no more than 75%. The business wants provisioning in 48 hours. You must decide whether to provision in SRP_PROD or request expansion. The project can accept phased delivery but not a performance SLA miss.
1. Can SRP_PROD safely accept the 40 TB thin request? Show the capacity reasoning.
Read the worked answer
Not as a single 40 TB request without reclaim or expansion. The SRP is already subscribed at 175 TB of 200 TB usable, leaving 25 TB of free capacity. Adding 40 TB of thin volumes takes subscribed capacity to 215 TB, which is over the usable capacity by 15 TB. Thin provisioning allows over-subscription, but that is a risk decision, not free space.
The mitigating data is real but not immediate: 18 TB of thin volumes have never been written, and compression is 1.6:1. Those factors may reduce physical consumption over time, but they do not create 15 TB of usable headroom today. The 30 TB of snapshot usage and the existing 25 TB free must also cover snapshot growth and any new writes. A phased 20 TB request still pushes subscribed to 195 TB, leaving only 5 TB free before snapshots and growth; that is too tight for a production SRP without a reclaim plan.
Recommendation: request expansion or reclaim at least 15-20 TB of committed capacity before approving the full request. If the business accepts phasing, start with 10-15 TB after reclaim and hold the remainder until free capacity is above an agreed threshold.
2. What performance checks would you complete before approving the 25,000 IOPS workload?
Read the worked answer
Capacity headroom is only half the approval. Check whether the array can absorb the additional 25,000 IOPS at the required response time, and whether the load lands on already-hot front-end directors.
Use Unisphere Performance and Workload Planner to review:
- Per-director and per-port front-end utilization, not just the 68% array peak. A new host zoned to one hot director could push that director over the 75% service level.
- Back-end utilization and response time at the 70k IOPS peak, plus cache dirty and write-pending behavior for a 70/30 read/write mix.
- Existing service levels by storage group; 25,000 IOPS is meaningful but may be small relative to 95k peak if distributed across ports.
- Workload Planner output for the proposed 40 TB and IOPS profile, including whether the planner assumes capacity that does not yet exist.
Approve performance only with a placement plan: distribute the new host's paths across port groups and directors, avoid a single hot director, and define a post-provisioning validation window. If the planner or per-director data shows a director crossing 75%, request rebalancing or expansion before go-live.
3. What actions would you take to free capacity and reduce risk before provisioning?
Read the worked answer
Build a reclaim and guardrail plan rather than relying on compression and thin over-subscription.
- Identify never-written or stale thin volumes and confirm with application owners whether they can be deleted or shrunk. Do not reclaim merely because a volume is thin; verify it is not a future allocation.
- Review SnapVX snapshots and linked targets. Expire snapshots beyond retention and terminate links that are no longer needed. Snapshot space is part of the capacity risk.
- Confirm compression is enabled where supported and appropriate, but do not count expected compression as immediate free capacity.
- Set SRP alerts at 75% and 85% used, and include snapshot and thin over-subscription in the report.
- Provision in phases with a gate: first phase only after free capacity is above the agreed threshold, second phase after 30 days of measured consumption.
- Re-run Workload Planner after reclaim so the performance decision uses current capacity and utilization, not the pre-reclaim state.