documentdb-storage

Files (5)

• SKILL.md
• storage-choose-premium-ssdv2.md
• storage-disk-hydration-sequencing.md
• storage-premium-ssdv2-limitations.md
• storage-tier-iops-caps.md

SKILL.md (first lines)

# Storage — Azure DocumentDB

Azure DocumentDB clusters use **remote premium SSD storage**. Two disk types are offered:

| Storage type | IOPS / bandwidth scaling | Max IOPS | CMK | Use it for |
|---|---|---|---|---|
| **Premium SSD** (v1) | Scales with **disk capacity** — bigger disk → higher IOPS | 20,000 | ✔️ | Legacy clusters, scenarios that require CMK |
| **Premium SSD v2** | **Decoupled** from disk size — IOPS/bandwidth gated by **compute tier** | 80,000 | ❌ | New production clusters; any I/O-intensive workload |

**Premium SSD v2 is the default recommendation for new clusters.** It delivers up to a **12× performance boost at no added cost** by removing the disk-size lever from the IOPS equation — you size storage for *capacity* and size compute for *throughput*, independently.

## How Premium SSD v2 changes sizing

On Premium SSD v1, hitting 20,000 IOPS required provisioning a 20 TB disk even when only ~1 TB of data was stored. On Premium SSD v2, the **highest achievable IOPS and bandwidth for the chosen compute tier are auto-configured** regardless of disk size:

- Choose storage size based on **how much data** the cluster will hold.
- Choose compute tier based on **how much IOPS / MBps** the workload needs (see [storage-tier-iops-caps](storage-tier-iops-caps.md)).
- No knobs to tune — the upper-bound IOPS/bandwidth for the tier are applied automatically at no extra cost.

## Rules

- [storage-choose-premium-ssdv2](storage-choose-premium-ssdv2.md) — When to pick Premium SSD v2 vs Premium SSD; the CMK trade-off.
- [storage-tier-iops-caps](storage-tier-iops-caps.md) — Compute-tier → max IOPS / MBps table; size compute for throughput, not disk for IOPS.
- [storage-premium-ssdv2-limitations](storage-premium-ssdv2-limitations.md) — Known limits: no CMK, capacity-change cap, replication restrictions, no online v1→v2 migration.
- [storage-disk-hydration-sequencing](storage-disk-hydration-sequencing.md) — Space out compute scale / storage scale / HA-enable operations while the disk is hydrating.

## References

- [Premium SSD v2 disks — Azure DocumentDB](https://learn.microsoft.com/azure/documentdb/high-performance-storage)

From Azure/documentdb-agent-kit, MIT.

More from Azure/documentdb-agent-kit