Pure Storage

---
name: pure-storage
description: "Pure Storage FlashArray (Everpure): Purity OS, DirectFlash modules, ActiveCluster, ActiveDR, SafeMode, Pure1 AIOps, Evergreen subscriptions, Pure Fusion, CSI driver, Cloud Block Store, and Portworx integration. Use for \"Pure Storage\", \"FlashArray\", \"Everpure\", \"Purity\", \"DirectFlash\", \"ActiveCluster\", \"ActiveDR\", \"SafeMode\", \"Pure1\", \"Pure Fusion\", \"Evergreen\", \"FlashArray//X\", \"FlashArray//C\", \"FlashArray//E\", \"Cloud Block Store\", \"Portworx\"."
license: MIT
---

# Pure Storage FlashArray

This skill covers Pure Storage FlashArray (marketed as Everpure since February 2026), including:

- FlashArray models: //X (performance), //XL (extreme scale), //C (capacity), //E (archival), //ST (SSD-based mid-market)
- Purity OS: always-on inline data reduction, global dedup, compression, pattern removal
- DirectFlash Modules (DFMs): proprietary flash media with no per-device controller/FTL
- ActiveCluster: synchronous active/active replication (zero RPO/RTO)
- ActiveDR: continuous async replication (near-zero RPO)
- SafeMode: immutable snapshots for ransomware protection
- Pure1: AIOps platform with Meta AI, Workload Planner, VM Analytics, Security Assessment
- Evergreen subscription model: Forever, Flex, One (STaaS with 99.9999% SLA)
- Pure Fusion: autonomous storage delivery control plane
- CSI Driver / PSO: Kubernetes dynamic provisioning for FlashArray and FlashBlade
- Cloud Block Store: Purity on AWS for hybrid cloud DR
- Portworx: Kubernetes-native storage and data management
- File services (NFS/SMB), S3 object protocol, vVols, NVMe-oF (RoCE/TCP)

For cross-platform storage questions, see the `overview` skill.

## How to Approach Tasks

1. **Classify** the request:
   - **Troubleshooting** -- Load `references/diagnostics.md` for Pure1 monitoring, performance analysis, connectivity issues, replication lag, hardware health, support workflow
   - **Architecture / design** -- Load `references/architecture.md` for model family, Purity OS, DirectFlash, ActiveCluster, ActiveDR, Pure1, Evergreen, Fusion, CSI, FlashBlade, Portworx
   - **Best practices** -- Load `references/best-practices.md` for volume provisioning, protection groups, SafeMode, ActiveCluster design, Kubernetes integration, performance optimization

2. **Identify context** -- Purity version (6.4.x vs 6.6.x), model generation (R4 vs R5), protocol (NVMe/TCP preferred from 6.4.2+).

3. **Apply Pure philosophy** -- Do not manually tune garbage collection, data reduction, or QoS unless isolation is specifically required. Purity manages these globally by design.

## Core Architecture

### DirectFlash Modules (DFMs)
Proprietary flash with no embedded controller, DRAM, or per-device FTL. All management (wear leveling, GC, bad block) done globally by Purity OS. Eliminates SSD "write cliff" and GC jitter. 75 TB and 150 TB shipping; 300 TB targeted. 39-54% fewer watts/TiB vs competitors.

### Always-On Data Reduction
Inline global dedup + LZ4 compression + zero-block pattern removal on every write. Cannot be disabled. Average 5:1 effective ratio. No performance-vs-efficiency trade-off.

### Model Selection

| Model | Use Case | Key Differentiator |
|---|---|---|
| //X | Mission-critical OLTP, VDI, databases | Full DFM NVMe, R5 = 30% perf boost |
| //XL | Extreme-scale mission-critical | Larger controller headroom |
| //C | Enterprise file, general, secondary | Cost-optimized, 16.3 PB effective |
| //E | Archival, cold flash | Lowest $/TB in family |
| //ST | Mid-market | SSD-based (no DFMs), 400 TB usable |

### Replication

| Method | Type | RPO | RTO |
|---|---|---|---|
| Protection Groups | Schedule-based async | Minutes | Minutes |
| ActiveDR | Continuous async | Seconds | Minutes |
| ActiveCluster | Synchronous active/active | Zero | Zero |
| Cloud Block Store | Async to cloud | Minutes | Minutes |

ActiveCluster: max 11 ms RTT, requires Pure1 Cloud Mediator or on-premises mediator, Uniform or Non-Uniform topology.

### SafeMode
Immutable snapshots locked for configurable period (24h default, up to 30 days). Only Pure Support can release after identity verification. Enable on all production arrays with minimum 72-hour eradication delay.

### Evergreen Model

| Tier | Ownership | SLA |
|---|---|---|
| Evergreen//Forever | Customer (perpetual + refresh) | Best-effort |
| Evergreen//Flex | Customer (term subscription) | Best-effort |
| Evergreen//One | Pure (STaaS) | 99.9999% + performance + energy |

## Top 10 Best Practices

1. Enable SafeMode on all production arrays (72h+ eradication delay)
2. Always configure multipath I/O — minimum 4 paths per host on FC
3. Set host personality on every host object
4. Group volumes by application consistency in protection groups
5. Deploy ActiveCluster with Pure1 Cloud Mediator on third network
6. Use NVMe/TCP for new Ethernet deployments (35% lower latency than iSCSI)
7. Set MTU 9000 end-to-end for iSCSI and NVMe/TCP
8. Set CSI StorageClass `reclaimPolicy: Retain` for production PVCs
9. Use Pure1 Workload Planner quarterly for capacity/performance forecasting
10. Do not manually tune GC, data reduction, or QoS — Purity manages globally

## Reference Files

- `references/architecture.md` -- Model family, Purity OS, DirectFlash, ActiveCluster, ActiveDR, Pure1 Meta AI, Evergreen, Pure Fusion, CSI driver, FlashBlade, Portworx, Cloud Block Store
- `references/best-practices.md` -- Volume provisioning, protection groups, SafeMode, ActiveCluster design, Kubernetes integration, performance optimization, capacity planning
- `references/diagnostics.md` -- Pure1 monitoring, performance troubleshooting, connectivity issues, replication lag, hardware health, support workflow, common issues