Microsoft Azure Cosmos DB Developer Specialty
225 practice questions
Last reviewed: April 2026
Personal notes and resource links for your study journey
Filter by Certification
DP-420 is Microsoft's specialty credential for developers building applications on Azure Cosmos DB for NoSQL. It validates the ability to design and implement data models, plan partition and distribution strategy, integrate Cosmos DB with surrounding Azure services, optimize performance and cost, and maintain Cosmos DB solutions in production. The audience is professional developers and data engineers who write Python, .NET, or JavaScript / TypeScript against the Cosmos DB SDK. The exam is heavily code-and-modeling focused: expect 40–60 questions in 100 minutes including code-completion drag-and-drops (SDK snippets, SQL API queries), scenario items, and at least one case study.
Largest domain at 37%. Document modeling for NoSQL workloads (denormalization, embedding vs. referencing), partition-key design, change-feed patterns, indexing policies (included / excluded paths, composite indexes, spatial indexes), and TTL configuration.
About 8%. Multi-region replication, multi-region writes, consistency-level tradeoffs (strong / bounded staleness / session / consistent prefix / eventual), conflict-resolution policies, and global distribution patterns.
About 8%. Change feed processor, Azure Functions Cosmos DB triggers, Event Hubs / Kafka integration, Cosmos DB analytical store with Azure Synapse Link, and integration with Azure AI Search.
About 17%. Request unit (RU) sizing and tuning, autoscale vs. manual throughput, indexing optimization, query performance, and cost analysis with the capacity calculator.
About 30%. Backup and restore (continuous and periodic), disaster recovery, security (Microsoft Entra auth, RBAC, customer-managed keys, IP firewall, Private Endpoint), monitoring (Azure Monitor, diagnostic logs), and SDK retry / error handling.
Services you'll encounter on the exam and why each one matters.
Native JSON document API (formerly SQL API) with SQL-like query syntax, server-side stored procedures, triggers, UDFs, and the canonical surface for DP-420 scenarios.
Why it's on the exam: Domain 1 (Design and implement data models) is dominated by NoSQL-API document modeling, partition-key choice, and embedded-vs-referenced trade-offs.
Wire-protocol-compatible MongoDB API on Cosmos DB infrastructure — supports vCore and RU-based modes plus standard Mongo drivers and aggregation pipeline operators.
Why it's on the exam: Domain 1 + Domain 3 (Integrate) test API selection — Mongo API is the migration-friendly choice when the application already speaks Mongo.
CQL-compatible wide-column API on Cosmos DB — keyspaces, tables, primary keys with partition + clustering columns, and Cassandra-driver compatibility.
Why it's on the exam: Domain 1 expects you to map Cassandra primary-key design (partition + clustering) onto Cosmos partitioning, and Domain 3 covers driver/migration scenarios.
Distributed Postgres (built on Citus) — shards relational tables across nodes via distribution columns, with reference tables and HTAP query routing.
Why it's on the exam: Domain 1 + Domain 2 (Distribution) test the relational/distributed trade-offs — Cosmos for PostgreSQL is the answer for SQL workloads that need horizontal scale.
Graph API on Cosmos DB — vertices and edges with property-graph traversals via Gremlin query language, with multi-region writes and tunable consistency.
Why it's on the exam: Domain 1 covers graph data modeling (vertex/edge cardinality, partition strategy for graphs) as the canonical use case for the Gremlin API.
Premium-tier key-value Table API surface — drop-in replacement for Azure Table Storage with global distribution, secondary indexes, and dedicated throughput.
Why it's on the exam: Domain 1 distinguishes the Table API from NoSQL for simple PartitionKey/RowKey workloads where a richer document model would be over-engineering.
Persistent ordered log of inserts and updates per logical partition, consumed via the Change Feed Processor library or Azure Functions Cosmos DB trigger.
Why it's on the exam: Domain 3 (Integrate) anchors event-driven integration on Change Feed — fan-out to materialized views, search indexers, and downstream services.
Native client libraries with bulk execution, automatic retry policies, point operations, transactional batches, LINQ + parameterized queries, and Direct-mode TCP transport.
Why it's on the exam: Domain 3 (Integrate) tests SDK patterns — request options, point reads vs queries, ConsistencyLevel overrides, and PartitionKey + ETag concurrency.
JavaScript server-side programs scoped to a single logical partition — transactional batch writes, pre/post triggers, and user-defined functions invocable from SQL queries.
Why it's on the exam: Domain 1 + Domain 4 (Optimize) test when to push logic server-side for atomicity vs latency — single-partition-scope is a recurring distractor.
High-throughput bulk import/update library that maximizes provisioned RUs by batching by partition, distributing across server-side connections, and back-pressuring on rate limits.
Why it's on the exam: Domain 3 + Domain 4 cite Bulk Executor for initial data migration and large reprocessing jobs — the canonical answer for "ingest millions of documents fast".
In-region read cache fronting Cosmos DB via a dedicated gateway — point-read and query caching with per-request `MaxIntegratedCacheStaleness` control.
Why it's on the exam: Domain 4 (Optimize) names Integrated Cache as the read-side optimization to cut RU/sec consumption on hot read paths without an external Redis layer.
Serverless binding that consumes the Change Feed at scale — checkpointed via a lease container, with input/output bindings for cross-container fan-out.
Why it's on the exam: Domain 3 (Integrate) tests Functions Cosmos DB trigger as the path of least resistance for event-driven projections off the Change Feed.
HTAP analytical store (column-oriented) auto-synced from the transactional store, queried from Synapse Spark / Serverless SQL without consuming transactional RUs.
Why it's on the exam: Domain 3 + Domain 4 expect Synapse Link when analytical queries would otherwise saturate the transactional workload — zero-ETL is the canonical answer.
Automatically maintained alternate-projection containers built off the Change Feed — pre-aggregate or re-partition data for query patterns that don't fit the base container.
Why it's on the exam: Domain 1 + Domain 4 cite materialized views as the answer when one container can't satisfy multiple access patterns without expensive cross-partition queries.
Per-container or shared-database throughput that scales between 10% and 100% of a configured maximum RU/sec, billed hourly at the peak seen each hour.
Why it's on the exam: Domain 4 (Optimize) contrasts manual vs autoscale provisioning for spiky workloads — autoscale wins when peak-to-trough ratio exceeds ~5×.
JSON-defined indexing surface — included/excluded paths, composite indexes, spatial indexes, and consistent vs lazy indexing modes for query/write trade-offs.
Why it's on the exam: Domain 4 (Optimize) heavily tests indexing policy tuning — excluding unused paths drives RU savings and composite indexes unlock ORDER BY queries.
Cloud identity directory; Cosmos DB control-plane and data-plane RBAC are bound to Entra principals via built-in and custom role definitions and role assignments.
Why it's on the exam: Domain 5 (Maintain) names Entra ID + Cosmos data-plane RBAC as the recommended path away from master-key authentication for production workloads.
Managed key store backing Cosmos DB customer-managed-key (CMK) encryption at rest, with key versioning, soft-delete, and managed-identity-based access.
Why it's on the exam: Domain 5 tests CMK rotation, double encryption (service-managed + CMK), and the operational impact of revoking a customer-managed key.
Telemetry pipeline for Cosmos DB — RU/sec, latency, normalized RU consumption, throttling, and diagnostic-log queries (DataPlaneRequests, QueryRuntimeStatistics) via KQL.
Why it's on the exam: Domain 5 (Maintain) is dominated by alerting on 429s, normalized RU consumption, and surfacing expensive queries from diagnostic logs.
Threat-detection layer that flags anomalous data-plane access, suspicious export patterns, and SQL-injection attempts against the Cosmos NoSQL API.
Why it's on the exam: Domain 5 references Defender for Cosmos DB as the security-monitoring complement to RBAC + network ACLs — required reading for any threat-detection scenario.
$110k–$150k–$210k USD annual
Range covers US-based mid-to-senior backend developers where Cosmos DB proficiency is required. Senior engineers building globally distributed apps at FAANG / fintech often clear $230k TC. Cosmos-DB-specific salary data is sparser than for general Azure roles given the narrower talent pool; figures lean on adjacent NoSQL / cloud-developer roles.
Source: levels.fyi 2025 backend / cloud developer roles, U.S. BLS OEWS May 2024 (15-1252 software developers, 15-1242 database administrators), Glassdoor 2025. Figures are approximate; actual compensation depends on role, region, and experience.
DP-420 sits in a narrower but well-paid niche — applications that genuinely need globally distributed, low-latency, multi-API NoSQL storage. Demand is concentrated at gaming companies, IoT platforms, retail / e-commerce at scale, and Microsoft-partner consultancies. Recruiters use it as a strong signal of deep Cosmos DB modeling and tuning competence, which commands a premium given the limited pool of qualified candidates. It pairs naturally with AZ-204 (Developer Associate) for full-stack Cosmos developers and with AI-102 / AI engineering roles where Cosmos DB serves vector and operational data for RAG architectures. Demand has been steady, with modest growth from Cosmos DB's expansion as a vector store for GenAI applications through 2024–2026.
There are no formal prerequisites. Microsoft recommends practitioner-level developer experience (one to two years professional development) plus working familiarity with Cosmos DB. Candidates without prior Cosmos DB exposure typically need significant additional time. AZ-900 and DP-900 are useful conceptual on-ramps for candidates new to Azure or to NoSQL data platforms; AZ-204 is highly complementary, given DP-420 assumes Azure-developer-level fluency with SDK patterns, Microsoft Entra authentication, and managed identities.
Proficiency in C#, Python, or JavaScript / TypeScript is essentially required: code-completion drag-and-drops show real Cosmos DB SDK snippets, with .NET examples most heavily represented in Microsoft's study material. The official Microsoft Learn path covers all five domains in roughly 30–40 hours. Hands-on time is essentially required — a personal Azure subscription with a small Cosmos DB account (or the Cosmos DB free tier) lets candidates practice partition-key design, indexing policies, and change-feed scenarios.
DP-420 sits in the Specialty tier and is broadly considered moderately to highly challenging — comparable to AZ-204 in code-completion difficulty, with narrower but deeper Cosmos-DB-specific surface area. Plan on 70–110 hours of study over 8–12 weeks for candidates with prior Cosmos DB experience; substantially longer otherwise. The exam runs about 100 minutes with 40–60 questions in multiple-choice, multiple-response, drag-and-drop (including code-completion), hot-area, and case-study formats. Case studies are timed separately and cannot be revisited.
The most common stumbling block is partition-key selection — the exam consistently presents nuanced workload patterns and expects candidates to identify the partition key that distributes load evenly while keeping common queries on a single partition. Indexing policy questions (included / excluded paths, composite indexes, query cost analysis) are another consistent trap area. As a Specialty exam, third-party study material is sparser; lean primarily on Microsoft Learn and the Cosmos DB documentation.
Most recent skills-measured update. Added vector-search coverage for AI workloads, expanded continuous-backup framing, modernized Microsoft Entra and customer-managed-key content. Microsoft refreshes DP-420 less frequently than role-based exams given its specialty status — typically every 18–24 months.
Restructured into the current five-domain layout, expanded change-feed and Synapse Link coverage, and integrated continuous-backup content.
Initial GA as Microsoft's first dedicated Cosmos DB developer credential. Original outline focused on the SQL (Core) API only and emphasized partitioning, RU sizing, and SDK patterns.
DP-420 (Microsoft Azure Cosmos DB Developer Specialty) is a a deeply specialized exam covering advanced topics in a narrow domain — expect hands-on experience to be a prerequisite Specialty-level exam. Most candidates need 100–200 hours of study spread over 2–4 months for specialty exams. These assume hands-on experience in the specialty domain. Most candidates who score consistently above the passing threshold on practice exams pass on their first attempt.
Most candidates need 100–200 hours of study spread over 2–4 months for specialty exams. These assume hands-on experience in the specialty domain. Time-to-pass varies widely by prior experience. Engineers with hands-on production experience in the underlying technology typically need less; candidates new to the platform should plan toward the upper end of that range.
DP-420 is a recognized credential in the Azure ecosystem and signals validated knowledge to employers, recruiters, and clients. Whether it is worth the time and fee for you depends on your role and goals — it tends to pay off most for cloud engineers, architects, and consultants who work with Azure day-to-day or want to move into roles that do.
The passing score for DP-420 is 700 / 1000. The exam contains 50 questions and lasts 1 hr 40 min.
The DP-420 exam fee is $165 USD. Fees are set by Azure and may vary by region; always confirm the current price on the official Azure certification page before booking.
Microsoft role-based certifications expire after 1 year but can be renewed for free via an unproctored online assessment on Microsoft Learn, starting 6 months before expiration.
Yes. You can take the exam online (proctored via the provider's secure browser, available 24/7 in most regions) or at an in-person Pearson VUE test center during business hours. Both formats use the same questions, time limit, and passing score.
CertLabPro provides 15 study modes across the practice question bank for DP-420. The exam-simulation mode mirrors the real exam: 50 questions in 1 hr 40 min, with the same passing threshold of 700 / 1000. Browse mode lets you read every Q&A statically.