Glossary
Enterprise flash storage
Enterprise flash storage is solid-state storage built for continuous data centre duty: drives and systems rated for round-the-clock operation, higher write endurance, protection against power loss and more consistent performance than flash made for personal computers.
Why enterprise flash matters in large storage systems
A laptop drive sees bursts of work separated by long idle periods, and its owner experiences an average. A drive in a storage cluster serves requests without pause, sits beside hundreds or thousands of identical drives, and is judged by its worst moments: a stall during background cleanup, a lost write after a power cut, a firmware hang that takes a node offline. At petabyte scale drive failures and replacements are routine operational events, and the features that separate enterprise from client flash determine how much each of those events costs in performance and risk.
Client and enterprise drive classes
The JEDEC JESD218 standard defines the conditions under which client and enterprise SSD endurance is rated. The two classes assume very different lives:
| Condition | Client class | Enterprise class |
|---|---|---|
| Active use | 8 hours a day at 40°C | 24 hours a day at 55°C |
| Powered-off data retention | 1 year at 30°C | 3 months at 40°C |
| Uncorrectable bit error rate | 1 in 1015 bits or better | 1 in 1016 bits or better |
The shorter retention figure for enterprise drives reflects their role: they are expected to stay powered, and their cells are rated for heavier writing at higher temperature.
Features that distinguish enterprise drives
- Power-loss protection: capacitors hold enough energy to flush the drive's DRAM buffer to flash if power fails, so a write the drive has acknowledged survives an outage. This also allows the drive to acknowledge writes from DRAM, which lowers write latency.
- End-to-end data protection: checksums carried along the internal path from interface to NAND and back, so corruption inside the drive is detected.
- Endurance grades: read-intensive models commonly rated around one drive write per day over five years, mixed-use models around three. The arithmetic is set out under flash storage endurance.
- Over-provisioning: spare flash hidden from the host, giving garbage collection room to work and steadying performance.
- Data centre form factors and interfaces: NVMe over PCIe in U.2 and EDSFF shapes such as E1.S and E3.S, with dual-port options for arrays in which two controllers reach the same drive.
- Self-encryption: data encrypted on the drive with keys that can be erased to retire it.
Performance consistency
Client drives are usually rated fresh out of the box, when every block is empty and writes go straight to clean pages. Enterprise drives are rated in steady state, after the drive has been filled and overwritten until garbage collection runs continuously, which is the condition a drive in a busy cluster lives in. Enterprise specifications also state latency at high percentiles, such as the 99.99th, describing how slow the slowest requests become while background work runs. The SNIA performance test specification formalises steady state by preconditioning a drive and recording results only once they stay within a narrow band over successive rounds. Peak IOPS describes the best case; percentile latency describes the experience of the applications on top.
What enterprise flash means for storage teams at scale
Drive class becomes a per-role decision. Large read-intensive drives suit capacity tiers holding data written once and read many times, while mixed-use drives suit metadata, journals and write-heavy services. Buying three drive writes a day for a write-once object tier pays for endurance that will never be used; buying one for a busy metadata service shortens the replacement cycle.
Tail latency compounds across a cluster. A request in a distributed or erasure-coded system that touches several drives completes when the slowest one answers. One drive model with erratic garbage collection therefore slows requests across every server it is fitted to, which makes consistency a stronger selection criterion than headline IOPS.
Capacity points reveal the endurance trade. The same NAND is often sold as a 3.84 TB read-intensive drive and a 3.2 TB mixed-use drive, the difference held back as spare area. Across a petabyte, choosing the mixed-use grade gives up roughly a sixth of raw flash in exchange for endurance and steadier writes.
Power events are fleet events. A failed power feed drops hundreds of drives at once, and power-loss protection is what keeps acknowledged writes and metadata intact across all of them.
Some array-era features matter less in scale-out designs. Dual porting exists so two controllers can share a drive; shared-nothing systems protect data across servers and do not use it. Fleet uniformity matters more: one drive model and firmware level across hundreds of nodes keeps behaviour predictable and qualification manageable.
Enterprise flash in Scality RING XP
RING is software, so its enterprise drives are the ones fitted to the x86 servers chosen for each deployment. The RING XP test configuration Scality published used Dell PowerEdge R7615 servers with AMD EPYC 9124 processors and 3.2 TB PCIe NVMe mixed-use drives, and recorded 511 microseconds per GET and 741 microseconds per PUT on 4 KB objects.