DP-420 study plan

Model a document schema for three distinct access patterns (read-heavy, write-heavy, aggregate-reporting) and defend each partition key choice

Compare single-partition, hierarchical (multi-level), and synthetic partition key strategies against a 20 GB / 100K RU workload

Provision a Cosmos DB for NoSQL account with a database and containers configured for both serverless and provisioned throughput

Choose consistency levels per container and document the latency, RU, and staleness trade-off for each choice

Implement create, read, update, patch, and delete operations with the .NET or JavaScript SDK against the containers built in week 1

Replace a multi-document write flow with a transactional batch and verify all-or-nothing behavior on failure

Author one stored procedure, one pre-trigger, and one user-defined function, then invoke each from the SDK

Implement a change feed processor that reacts to inserts and updates and writes a projection to a second container

Configure a container with a hierarchical partition key and query it at both tenant and tenant-plus-user scope

Tune an indexing policy (included, excluded, composite, spatial) to cut RU charge on a measured query by at least 40 percent

Read change feed in all-versions-and-deletes mode and reconcile deletes into the projection container from week 2

Measure RU charge per operation for point read versus query versus cross-partition query and record the results

Harden the week 1-3 solution by replacing master keys with Microsoft Entra ID and role-based access control in application code

Add a private endpoint or service endpoint plus firewall rules and confirm the account rejects unauthorized network access

Implement optimistic concurrency with ETags and prove a conflicting update is rejected

Complete a timed 40-question practice set weighted 50 percent domain 1 and 50 percent domain 2 and log every miss by sub-topic

Configure autoscale versus standard throughput and select a partition key that avoids hot partitions under a simulated spike

Implement a retry and fallback policy for 429 responses and demonstrate the app sustaining load after throttling

Design a key rotation procedure for account keys and execute it against the dev account without downtime

Enforce data encryption expectations by documenting how service-managed and customer-managed keys differ in setup and operation

Deploy the full solution with Bicep or Terraform from a CI/CD pipeline in GitHub Actions or Azure DevOps

Parameterize environment-specific settings and run the pipeline against both a dev and a test environment

Configure Azure Monitor alerts for normalized RU consumption and server-side latency, and send them to an action group

Automate a database-level throughput or indexing policy change through the pipeline and confirm the change landed

Enable the Cosmos DB vector search capability and create a vector embedding policy plus a vector index on a container

Generate embeddings for a sample document set, store them, and run a top-k similarity query filtered by a partition key

Build a retrieval-augmented generation flow that queries Cosmos DB and passes results to Azure OpenAI

Make the RAG flow return grounded results with the source documents for each answer

Connect analytical workloads using Azure Synapse Link or the analytical store and query operational data without affecting transactional RU

Build a data movement path from Cosmos DB change feed into an analytics sink

Re-run two timed practice sets and target 85 percent or better weighted across all three domains

Review every remaining flagged sub-topic and close the gaps with handwritten summaries