GFS Retention on a 192 TB Dedupe Appliance
A regional hospital protects 500 VMware VMs totaling 80 TB with PowerProtect Data Manager writing to a PowerProtect DD appliance with 192 TB raw capacity. The current policy takes a daily incremental with a 3% change rate and a weekly full; all backups are kept 30 days. The compliance office now requires 12 monthly and 7 yearly copies, with the yearly copies immutable for 7 years. The backup team assumes a 5:1 deduplication ratio and wants to know whether the existing appliance can hold the new GFS policy. The appliance is already 55% used, and the team wants to keep at least 20% free for rehydration and maintenance. Cloud tiering to an object store is approved, but the CISO is worried about egress charges and restore time for a compliance request.
1. Calculate the logical and physical capacity demand of the new policy and decide whether the appliance suffices. Show your assumptions.
Read the worked answer
Logical demand under the stated assumptions: 30 daily incrementals at 3% of 80 TB = 2.4 TB each = 72 TB. Twelve monthly fulls at 80 TB = 960 TB. Seven yearly fulls at 80 TB = 560 TB. Total logical = 1,592 TB, about 1.59 PB. At the assumed 5:1 dedupe ratio, physical demand is about 318 TB. The appliance has 192 TB raw; with 20% kept free, usable is about 154 TB, and 55% is already used, leaving only about 48 TB. Even allowing that monthly and yearly fulls dedupe against each other better than 5:1, the new policy does not fit. The assumption that matters most is dedupe ratio: a 10:1 ratio would give about 159 TB, which is still at or beyond the usable limit once metadata and growth are included. The decision should be no for all retention on the appliance; the design must tier long-term copies or add capacity. I would not approve the policy based on a best-case dedupe assumption without a measured ratio from the existing data.
2. Design the GFS placement and retention-lock strategy. Address egress and restore time for a compliance request.
Read the worked answer
Keep the operational recovery window on the appliance: 30 daily incrementals plus the three most recent monthlies. That is about 72 TB + 240 TB = 312 TB logical, or roughly 62 TB at 5:1, which fits the free space with margin. Tier the remaining nine monthlies and all seven yearlies to the approved object store with retention lock in compliance mode for the required period. Apply immutability at the object store and keep the appliance copy unlocked only for the short-term window. For egress and restore, distinguish a compliance request for a few files from a full-site recovery. A file-level request can recall only the needed objects; a full 80 TB monthly restore over the WAN is not an acceptable RTO. Therefore keep at least the latest monthly and the latest yearly restorable locally, or maintain a staged recall process with a defined cache. Test a sample recall and measure time and egress. Do not promise 2-hour recovery from cloud tier for an 80 TB full; state the recovery tier explicitly in the service catalog.
3. What would you monitor after implementation to prove the design remains compliant and affordable?
Read the worked answer
Monitor logical versus physical capacity and dedupe ratio by policy, because a change in VM change rate or an application upgrade can break the sizing assumption. Track appliance free space against the 20% reserve, and alert before rehydration or maintenance is constrained. Monitor cloud tier recall latency, egress volume, and cost per restore request; a compliance request that recalls an entire yearly full can be unexpectedly expensive. Verify retention-lock status and expiry dates for every yearly copy, and alert on any copy that is not immutable. Monitor backup success by SLA, synthetic full validation, and restore-test results. Also monitor replication or tiering job backlog, because a lagging tier job can leave the appliance holding data longer than designed. Report these metrics monthly to compliance and finance; compliance needs evidence of immutability, and finance needs evidence that egress is controlled. A design that is compliant but unaffordable will not survive an audit cycle.