Tiering Media Archive and VMware Protection
A post-production company stores 280 TB of active media on an aging NAS and 18 TB of VMware VMs on an EMC VNX. Editors need predictable low-latency access to 40 TB of current projects; finished projects older than 90 days are rarely accessed but must be retained for seven years. Tier-1 VMware RTO is 15 minutes; other systems can tolerate four hours. Backup currently goes to tape with a 12-hour window and monthly fulls; last quarter's restore test failed. The company has 10 GbE and a DR site 30 km away but no dark fiber. The CFO wants one throat to choke but refuses all-flash for the archive. The IT director asks whether HPE Primera, Nimble, MSA, and StoreOnce can replace the estate without overbuying. A 300 TB usable target is forecast in three years.
1. Map the workloads to HPE storage products and justify the tiering decisions.
Read the worked answer
Map by SLA and IO, not by product tier:
- Tier-1 VMware, 18 TB: HPE Primera all-flash with synchronous or near-sync replication to DR if the 15-minute RTO requires it. This is the latency- and availability-sensitive tier; overbuying here is justified.
- Active media, 40 TB: HPE Nimble all-flash or hybrid, depending on measured throughput and editor concurrency. Validate streaming read performance and 10 GbE limits before choosing hybrid.
- Cold archive, remaining ~240-280 TB: HPE MSA with large LFF drives is the cost option. It is not an all-flash replacement, and it is acceptable only if integrity, snapshots, and backup meet the seven-year retention requirement. If not, use a higher-tier hybrid.
- Backup: HPE StoreOnce for VMware and metadata, with dedupe and replication to the 30 km DR site. StoreOnce is a backup target, not the primary archive. Do not all-flash the archive; the SLA does not justify it. Single throat to choke can be one support contract, not one product.
2. How would you size backup and retention, and what would you fix first after the failed restore test?
Read the worked answer
Resize backup around change rate and restore objectives:
- Move from monthly fulls to incremental-forever with StoreOnce Catalyst and replication.
- Keep 30-60 days local for operational restores, and replicate to DR for the four-hour tier. The 15-minute tier needs storage replication or snapshots, not just backup.
- Do not size StoreOnce at 300 TB raw. Estimate 18 TB VMware plus active media metadata, apply dedupe/compression assumptions, and validate with a proof. Keep 20-30% headroom.
- Fix the restore failure before buying: define restore tests per quarter, document RTO/RPO, and make application owners sign off.
- For seven-year archive, use the archive tier with integrity checks and a media migration plan, not backup tapes alone. The reasoning: backup capacity is driven by change rate and retention, while RTO is driven by recovery method. A bigger StoreOnce will not fix a missing DR runbook.
3. The CFO wants one throat to choke. How do you respond without forcing the wrong product on every workload?
Read the worked answer
Agree to a single accountable support model, not a single product. Propose HPE Pointnext as the support integrator across Primera, Nimble, MSA, and StoreOnce, with named service levels. Explain that forcing all data onto one platform would either overbuy all-flash for cold media or underserve tier-1 VMware. Use a reference architecture and a single escalation path. Success criteria: one case owner, one monthly service review, tested DR, and no finger-pointing between storage and backup teams. If the customer insists on one product, qualify the SLA trade-off in writing. The judgement is to sell the outcome the CFO wants, accountability, without pretending that one array fits all four data classes.