DP-420: Azure Cosmos DB Developer Specialty

DP-420: Designing and Implementing Cloud-Native Applications Using Microsoft Azure Cosmos DB — Key terminology reference for exam preparation.

Core Concepts

Azure Cosmos DB

Microsoft's globally distributed, multi-model NoSQL database service for cloud-native apps.

Request Unit (RU)

Normalized currency measuring the cost of database operations, independent of hardware.

Container

Fundamental unit of scalability/storage in Cosmos DB for NoSQL, analogous to a table.

Item (Document)

Fundamental unit of data, stored as a self-describing JSON document within a container.

API for NoSQL

Cosmos DB's native API (formerly SQL API), queried with a SQL-like language; DP-420's focus.

OLTP

Online Transactional Processing; small, fast operations — Cosmos DB's core transactional workload type.

OLAP

Online Analytical Processing; large aggregations and analytical queries over historical data

Data Modeling & Partitioning

Partition Key

Property used to distribute items across physical partitions for scalability.

Logical Partition

Group of items sharing the same partition key value; max 20 GB storage.

Physical Partition

Underlying infrastructure unit hosting one or more logical partitions.

Synthetic Partition Key

Concatenated/computed key combining multiple properties for better distribution.

Hierarchical Partition Key

Up to 3 levels of partition keys for granular, multi-tenant partitioning.

Embedding (Denormalization)

Nesting related data within a single document for fewer round trips.

Referencing (Normalization)

Storing related entities as separate documents linked by an id reference.

Time to Live (TTL)

Container/item setting that automatically deletes items after a set time.

Unique Keys

Container-level constraint enforcing uniqueness of property paths per logical partition.

Throughput & Scaling

Provisioned Throughput

Reserved, guaranteed RU/s billed regardless of consumption; suits steady workloads.

Serverless

Consumption-based billing for actual RUs used; suits intermittent/unpredictable traffic.

Autoscale Throughput

Automatically scales RU/s between 10%-100% of a configured maximum based on demand.

Database-Level Throughput

Provisioned throughput shared across all containers in a database.

az cosmosdb sql container update

CLI: updates container config (indexing, TTL, unique keys), not throughput.

az cosmosdb sql container throughput update

CLI: changes a container's manual RU/s or autoscale max RU/s.

az cosmosdb sql container throughput migrate

CLI: switches a container between manual and autoscale throughput.

az cosmosdb sql container redistribute-partition-throughput

CLI: reallocates RU/s unevenly across physical partitions to relieve a hot partition.

Data Distribution & Consistency

Global Distribution

Turnkey replication of data across any number of Azure regions, no code changes.

Consistency Levels

Five tunable levels (Strong, Bounded Staleness, Session, Consistent Prefix, Eventual).

Automatic Failover

Policy that automatically promotes a new write region during a regional outage.

Multi-Region Writes

Every region accepts reads and writes simultaneously (multi-master).

Conflict Resolution Policy

Defines how concurrent multi-region write conflicts are resolved (e.g., LWW).

Session Token

Value guaranteeing a client reads its own writes under session consistency.

Client Connectivity & SDK

Gateway Mode

SDK connectivity via a single HTTPS endpoint; more firewall-friendly, higher latency.

Direct Mode

SDK connectivity direct to backend replicas over TCP; lower latency, more ports needed.

Cosmos DB Emulator

Local, offline tool emulating Cosmos DB for development/testing without Azure cost.

Point Operation

Single-item operation using id + partition key, more RU-efficient than a query.

Patch Operation

Partial update of specific item properties without replacing the whole document.

Transactional Batch

Groups multiple point operations on the same logical partition into one atomic transaction.

Bulk Support

SDK feature optimized for loading/updating large volumes of items across many partitions.

Optimistic Concurrency (ETag)

Uses the ETag property with if-match preconditions to detect write conflicts.

Continuation Token

Opaque value used to resume a paginated query/feed from where the last page left off.

Cross-Partition Query

Query spanning multiple physical partitions; tuned via MaxConcurrency and MaxBufferedItemCount.

429 Throttling Error

Error indicating a request was rate-limited due to exceeding provisioned RU/s.

isTombstone

Server-Side Programming

x-ms-retry-after-ms

http header value that represents in milliseconds the minimal waiting time between requests.

Item response, Feed response

Contains both returned data plus request charge, session token, and ETag

Queries

Queries

Stored Procedure

JavaScript logic executed server-side within a single logical partition transaction.

BulkImportResponse

Return object of the classic bulkImport stored procedure pattern; reports numDocsImported/continuation.

Trigger

Pre- or post-operation JavaScript logic, explicitly invoked by the client per request.

User-Defined Function (UDF)

Read-only JavaScript function extending SQL queries with custom computations.

Aggregate Functions (COUNT, SUM, AVG, MIN, MAX)

Built-in SQL aggregate operators (COUNT, SUM, AVG, MIN, MAX) used with GROUP BY.

EXISTS & Array Functions (IS_ARRAY, ARRAY_CONTAINS)

SQL subquery and array functions (EXISTS, IS_ARRAY, ARRAY_CONTAINS, ARRAY_LENGTH) for nested data.

Integration & Analytics

Change Feed

Persistent, ordered log of item inserts/updates, consumable by apps and Azure Functions.

Full-Text Query

Native keyword search via FULLTEXTCONTAINS/FULLTEXTSCORE, backed by a full-text index.

HTAP (Hybrid Transactional and Analytical Processing)

Architecture running analytics over operational data without impacting transactional OLTP RUs, enabled via Synapse Link/analytical store.

Azure Synapse Link / Analytical Store

Near real-time analytics via a separate column-store, no RU impact on transactional data.

Azure Cosmos DB Mirroring for Fabric

Near real-time replication of Cosmos DB data into Microsoft Fabric's OneLake.

OneLake

Fabric's single, unified, tenant-wide data lake storing all workloads' data in Delta format.

Delta Format

Open-source ACID table format (Delta Lake) used for all data stored in OneLake.

Apache Spark

Distributed processing engine used in Synapse/Fabric notebooks to analyze Cosmos DB data.

Serverless SQL Pool

On-demand T-SQL query service in Azure Synapse to query Cosmos DB analytical store via OPENROWSET.

Direct Lake

Power BI mode reading Delta tables directly from OneLake — near-Import speed, always fresh.

Azure Functions Integration

Serverless compute triggered by Cosmos DB change feed for event-driven processing, denormalization, and data movement.

Azure Logic Apps

Low-code cloud workflow automation service integrating with Cosmos DB via connectors, triggers, and actions.

Azure AI Search Integration

Indexer pulls Cosmos DB data (via change feed) to enable full-text/semantic search.

Optimization

Indexing Policy

Defines which properties are indexed, trading off query performance vs. write RU cost.

Range Index

Default index type for string/numeric properties; supports equality, range, and ORDER BY.

Spatial Index

Accelerates geospatial queries (ST_DISTANCE, ST_WITHIN) over GeoJSON properties.

Composite Index

Index across multiple properties to support efficient multi-property filter/sort queries.

Vector Index

Enables approximate nearest-neighbor similarity search over embeddings for AI/RAG scenarios.

Wildcard / Included & Excluded Paths

Path expressions (with optional /* wildcard) controlling exactly which properties are indexed.

Integrated Cache

In-memory caching layer (via dedicated gateway) reducing RU cost/latency for repeated reads.

Change Feed Estimator

Reports pending change feed lag to help decide when to scale out consumer instances.

Maintenance & Security

Alerts

Cosmo DB alerts

Normalized RU Consumption

Azure Monitor metric showing % of provisioned throughput used; flags hot partitions.

PartitionKeyRUConsumption (Diagnostic Log)

Per-second RU/s consumption per partition key; pinpoints hot partitions.

PartitionKeyStatistics (Diagnostic Log)

Estimated storage size (KB) per logical partition key; flags storage skew.

QueryRuntimeStatistics (Diagnostic Log)

Logs query operations against the account for query performance analysis.

ControlPlaneRequests (Diagnostic Log)

Logs account control-plane operations, e.g. adding/removing replication regions.

DataPlaneRequests (Diagnostic Log)

Logs item-level create/read/replace/delete/query operations with RU charge and duration.

KQL (Kusto Query Language)

Query language used in Azure Monitor Log Analytics to analyze Cosmos DB diagnostic logs and metrics.

Periodic Backup

Original backup mode; scheduled snapshots, typically restored via a support request.

Continuous Backup & Point-in-Time Restore

Continuous backups enabling self-service restore to any timestamp in the retention window.

Customer-Managed Keys

Encrypt data at rest using a key you control in Azure Key Vault.

Data Encryption Key (DEK)

for Always Encrypted in Azure Cosmos DB, you must use the Azure Cosmos DB SDK (C#, Java, or Node.js).

RBAC for Data Plane

Fine-grained Microsoft Entra ID permissions for data operations, avoiding shared account keys.

Cosmos DB Operator

Provisions accounts, databases, and containers; cannot access data or use Data Explorer.

Cosmos DB Account Reader Role

Can read Azure Cosmos DB account data; cannot provision resources or make changes.

Cosmos Backup Operator

Submits a restore request for a periodic backup-enabled database or container.

CosmosRestoreOperator

Performs the restore action for an account in continuous backup mode.

Always Encrypted

Client-side encryption ensuring Cosmos DB never sees plaintext values of protected properties.

Azure Data Factory (Data Movement)

Pipeline-based ETL/ELT service for moving data into/out of Cosmos DB.

ARM Templates (Infrastructure as Code)

Declarative, version-controlled provisioning of Cosmos DB resources as code.

log-analytics

---

DP-420: Designing and Implementing Cloud-Native Applications Using Microsoft Azure Cosmos DB — terminology reference