Storage refresh
Your array is aging out: support renewal is climbing, capacity is tight, and the latency graphs are drifting the wrong way. A refresh decision comes down to fewer variables than the vendor decks suggest: how much usable capacity you need at what latency, which protocols your hosts actually speak, and whether you want a self-contained appliance or an architecture that scales controllers and capacity separately.
All-NVMe is the default now, not the premium option. Data-reduction ratios in vendor marketing assume friendly data; size against raw capacity and treat reduction as upside.
What actually determines the answer
| Factor | Why it matters |
|---|---|
| Usable capacity, honestly counted | Size on raw NVMe and protection overhead. Treat dedupe and compression claims as upside, not baseline. |
| Protocol reality | What your hosts speak today (FC, iSCSI, NVMe-TCP) constrains the shortlist more than performance claims do. |
| Appliance vs disaggregated | Appliances (PowerStore) are simple and predictable. Disaggregated (Alletra MP) scales performance and capacity independently, which pays off in larger or lopsided estates. |
| Migration path | Host-based migration works everywhere but takes windows. Array-assisted migration shortens the project when source and destination cooperate. |
The paths, with their trade-offs
Dell PowerStore appliance
When you want unified block and file in a predictable, self-contained 2U building block.
Trade-off: Scaling is appliance-sized steps; very large estates end up managing a cluster of them.
Configured BOM
PowerStore 1200T - virtualization array
- Appliance PowerStore 1200T, dual-node 2U base enclosure
- Media 12× 3.84 TB NVMe SSD
- Connectivity 25 GbE iSCSI / NVMe-TCP
- Support 4-year ProSupport Plus, next business day
HPE Alletra Storage MP
When block-only is fine and you want controllers and capacity to scale independently.
Trade-off: Block-focused; if you need file services too, that's a separate answer.
Configured BOM
Alletra MP B10000 - block storage building block
- Chassis 2U controller chassis, dual nodes
- Media 12× NVMe SSD, sized to your capacity target
- Management HPE GreenLake cloud console
- Support 4-year HPE support, next business day
What deployment involves
Typical scope: array install and configuration, fabric or IP storage network changes, host connectivity, and data migration by volume or datastore.
Most mid-sized migrations complete inside a month of maintenance windows; the array install itself is days.