Skip to main content

documentdb-storage

Storage configuration guidance for Azure DocumentDB — when and how to use Premium SSD v2 high-performance storage, IOPS/bandwidth caps that are gated by compute tier (not disk size), Premium SSD v2 limitations (no CMK, migration paths, disk-hydration sequencing), and storage capacity change limits. Use when picking a storage type at cluster creation, sizing for I/O-intensive workloads, migrating from Premium SSD to Premium SSD v2, or sequencing compute/storage/HA changes on a Premium SSD v2 cluster.

معلومات المصدر

المستودع
Azure/documentdb-agent-kit
آخر نشاط في المصدر
١٢ مايو ٢٠٢٦ في ١٧:٠١
لغة SKILL.md المكتشفة
الإنجليزية
النجوم
٦
التفرعات
١١

خيارات التثبيت

يُحدَّد Prompt الذي يراجع المصدر أولًا بشكل افتراضي. يمكنك التبديل إلى أمر مباشر أو تنزيل نسخة محلية.

مراجعة ملفات المصدر

اقرأ SKILL.md وأي ملفات مرافقة يعرضها SkillsMP قبل أن تقرر التثبيت.

مستكشف الملفات
5 ملفات

عرض SKILL.md

SKILL.md
تعليمات المصدر · معاينة للقراءة فقط
name
documentdb-storage
description
Storage configuration guidance for Azure DocumentDB — when and how to use Premium SSD v2 high-performance storage, IOPS/bandwidth caps that are gated by compute tier (not disk size), Premium SSD v2 limitations (no CMK, migration paths, disk-hydration sequencing), and storage capacity change limits. Use when picking a storage type at cluster creation, sizing for I/O-intensive workloads, migrating from Premium SSD to Premium SSD v2, or sequencing compute/storage/HA changes on a Premium SSD v2 cluster.
license
MIT
# 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)
عرض على GitHub