⚡ Official blueprint — 100-minute assessment; three equally important domainsMicrosoft does not promise a fixed question count or a guaranteed case-study count. Prepare for scenario questions and possible interactive components. Click each domain to expand.
Do not attempt every card tonight. Your target is tool selection, architecture reasoning, monitoring/error diagnosis, and code recognition. The quiz randomizes choices each run, so the score now measures knowledge rather than answer-position memory.
FIRST · 45 MIN
1. Architecture choices
- Pipeline = batch movement + orchestration
- Dataflow Gen2 = visual Power Query transformation
- Notebook/Spark = large or complex transformation
- Lakehouse = open files/Delta + Spark
- Warehouse = full relational T-SQL
- Eventstream → Eventhouse → Dashboard/Activator
NEXT · 90 MIN
2. Ingest and transform
- Full vs incremental, CDC and watermarks
- Shortcut vs mirroring vs Copy
- SCD1 vs SCD2; facts vs dimensions
- Recognize PySpark, SQL, KQL and pipeline expressions
- Duplicates, nulls and late events
DO NOT SKIP · 90 MIN
3. Monitor and optimize
- Monitoring Hub → item-native diagnostic surface
- Pipeline activity output; Dataflow refresh history
- Spark UI; DMVs vs Query Insights
- Workspace Monitoring/Eventhouse logs
- OPTIMIZE vs VACUUM; cache, partition and joins
60 MIN
4. Security and lifecycle
- Workspace role vs item permission vs data permission
- RLS, CLS/OLS, DDM and OneLake security
- Git = version control; deployment pipeline = promotion
- Deployment rules for environment bindings
- Least privilege: grant only the access path required
45–60 MIN
5. Scenario drill
- Batch or streaming?
- Copy or no-copy?
- SQL, Spark, Power Query or KQL?
- Primary constraint: latency, cost, security or maintenance?
- Which minimum set of items satisfies every requirement?
FINAL · 30 MIN
6. Exam execution
- 100 minutes: answer every question; no penalty for guessing
- Read section instructions—after leaving some sections you cannot return
- Microsoft Learn is available on associate/expert exams, but the timer continues
- Use Learn for rare syntax/limits, not for every question
- Review the official practice assessment and exam sandbox
Use your SAP BW/HANA backgroundFabric pipeline ≈ process-chain/orchestration thinking; incremental/CDC ≈ delta loads; Bronze/Silver/Gold ≈ raw/staging/curated layers; Warehouse DMVs/Query Insights ≈ live and historical HANA workload diagnosis.
🗺️ How all Fabric components connect — click any colored box to open its Concept cardData flows top-to-bottom: Sources → Ingestion → OneLake storage → Compute/Process → Serving → Consumption. Governance/Operations wrap every layer.
Data Factory / Pipeline / Warehouse / OneLake
Lakehouse / Dataflow / Semantic Model
Spark / Delta Lake
Real-Time / KQL / Eventhouse
Shortcut / Activator / DW Security / Admin
Click any colored box → opens Concept Explorer card for that component
Official Microsoft visual references — recreated as exam study diagrams
Fabric SaaS architecture
Workloads sit on a shared Fabric platform. OneLake is the storage foundation, while governance, security, monitoring, and Copilot are shared services.
Experiences: Data Factory · Data Engineering · Data Warehouse · Real-Time · Power BI
Shared platform: OneLake · security · governance · monitoring · capacity
Exam rule: choose the workload by scenario, but remember they share OneLake.
OneLake hierarchy
The hierarchy is tenant-wide, not capacity-wide. Most Fabric data items live under workspaces and use the same OneLake namespace.
Tenant: one unified OneLake
Workspaces: logical containers for Fabric items
Items: Lakehouses · Warehouses · KQL databases · semantic models
Lakehouse end-to-end
DP-700 scenario questions often ask where work should happen: ingestion, raw storage, transformation, curated serving, or monitoring.
Get data: Pipeline · Dataflow Gen2 · Shortcut · Eventstream
Transform: Spark notebooks · SQL · Power Query
Serve: SQL endpoint · semantic model · reports · dashboards
| Visual pattern | What to memorize for DP-700 | Trap answer to avoid |
|---|---|---|
| Fabric architecture | Different workloads, one SaaS platform, shared OneLake and governance. | Copying data between every engine by default. |
| OneLake hierarchy | Tenant → Workspace → Item → Files/Tables. OneLake is one per tenant. | OneLake per capacity or per workspace. |
| Lakehouse architecture | Bronze/Silver/Gold layers normally separate raw, cleansed, and curated data. | Transforming heavily before Bronze when the requirement says raw landing. |
🧠 Concept Explorer — understand the "what", "why", and "how it connects" for every componentSearch by keyword or click a category tag. Click "Connects to" pills to jump between related concepts.
Click or press Space to reveal answer
1 / ?
Missed: 0 | Got it: 0
← Prev Space Flip → Next G Got it M Missed
💻 Language-organized syntax reference for DP-700 — every code pattern the exam expects you to recognize or write. Not memorization — pattern recognition.
🎯 DP-700 scenario training + Contoso case-study lab. These build business-context reasoning across several requirements. Microsoft does not guarantee a fixed case-study or question count. If the delivered exam includes a section that cannot be revisited, the opening instructions will say so—review before leaving it.
🧩 Contoso case-study lab — converted from the supplied practice materialRead the tabs/requirements before answering. In the real exam, review every answer before leaving a locked case-study section. Repeated questions were consolidated here, and missing screenshot exhibits were converted into explicit code and decision tables.
1. Case facts at a glance
| Area | Current state | Exam consequence |
|---|---|---|
| Capacity | F64 Cap1; WorkspaceA and WorkspaceB currently use Pro license mode | Fabric items must be assigned to suitable Fabric capacity before the solution runs there. |
| Teams | Data engineers prefer Python/SQL. Data analysts use Power Query and T-SQL. | Favor notebooks/pipelines for engineering and SQL/Dataflow/semantic experiences for analysts. |
| POS1 | SQL Server on an Azure VM, private VNet, public access blocked | Use private connectivity. The appropriate component depends on the workload: gateway for pipeline copy; managed private endpoint for Spark access to a private Azure endpoint. |
| MAR1 | SaaS marketing app with seven REST endpoints and transient connection failures | Use a built-in REST connection, parallel ingestion where possible, and retry policies. |
| History | 12 Parquet files in Amazon S3, each 300–900 MB | Use an S3 shortcut and shortcut cache: no persistent raw copy and reduced repeated cross-cloud egress. |
| Product hierarchy | Products → ProductSubcategories → ProductCategories | Filter active products first, then join outward so unassigned categories/subcategories are naturally excluded. |
2. Target architecture
SourcesPOS1 private SQL
MAR1 REST APIs
S3 Parquet history
MAR1 REST APIs
S3 Parquet history
IngestionPipelines + retry
Parallel REST copies
S3 shortcut/cache
Parallel REST copies
S3 shortcut/cache
Lakehouse1Bronze: raw Delta
Silver: cleansed and standardized
Silver: cleansed and standardized
Lakehouse2Gold dimensional model
Active products only
Active products only
ConsumptionWorkspaceB
Semantic models, reports, dataflows
Semantic models, reports, dataflows
| Requirement | Design response |
|---|---|
| Complete each layer before the next | Pipeline dependencies: Bronze success → Silver; Silver success → Gold. |
| Email on any failure | Connect each failure path to notification logic; preserve the failed activity output for diagnosis. |
| Run imports simultaneously | Use parallel pipeline branches or a ForEach with controlled concurrency. |
| No pre-Bronze transformation except format | Land source-shaped data as Delta; perform deduplication, missing-value handling, and capitalization standards in Silver. |
| Remove obsolete Delta files weekly | Schedule VACUUM with the organization-approved retention interval; understand that VACUUM removes files, while OPTIMIZE compacts them. |
| Source control | Store Lakehouses, pipelines, and notebooks in WorkspaceA and commit supported item definitions to Azure Repos. Data, credentials, and refresh schedules need separate handling. |
3. Security and workspace placement
| Group | Needed access | Least-privilege design |
|---|---|---|
| DataEngineers | Read/write all Lakehouses and underlying files; commit WorkspaceA changes | WorkspaceA Contributor or higher as required for the specific item operation, plus source-control permissions. |
| DataAnalysts | Read only Gold Delta tables; no Bronze/Silver access | Share Lakehouse2 directly and grant SQL endpoint data read. Do not add them to WorkspaceA, because a workspace role broadens visibility. |
| Restricted writer | Write Customer without reading Employee PII | Separate the PII table into another Lakehouse/workspace, then grant write access only in the non-PII workspace. Workspace Contributor is not table-scoped. |
4. Recreated exhibits: code and tables
Year/product totals plus a year-level subtotal
SELECT
YEAR(o.OrderDate) AS OrderYear,
p.ProductName,
SUM(o.OrderQuantity) AS TotalOrderQuantity
FROM dbo.Orders AS o
JOIN dbo.Products AS p
ON o.ProductID = p.ProductID
GROUP BY GROUPING SETS
(
(YEAR(o.OrderDate), p.ProductName),
(YEAR(o.OrderDate))
);Why GROUPING SETS? It returns product totals within each year and an additional all-products subtotal for each year, without forcing an unwanted grand-total row.
Flat-table columns: fact or product dimension?
| Column | Destination | Reason |
|---|---|---|
| ProductID | DimProduct | Natural/business key identifying the product. |
| ProductName | DimProduct | Descriptive product attribute. |
| ProductColor | DimProduct | Descriptive product attribute whose changes can be tracked. |
| Date | FactSales / DimDate key | Describes when the transaction occurred, not the product. |
| TransactionID | FactSales | Transaction identifier or degenerate dimension. |
| SalesAmount | FactSales | Numeric measure to aggregate. |
Deployment behavior
| Artifact | What moves | What needs separate handling |
|---|---|---|
| Reports / semantic models | Supported definitions | Refresh schedules and environment bindings |
| Notebooks / pipelines | Supported definitions | Connections, credentials, schedules, and target references |
| Lakehouse | Item definition/stub | Tables, files, and data population |
| Warehouse | Schema definition through supported deployment paths | Data and credentials; validate references with a SQL database project build |
| Eventhouse | Supported by Fabric deployment pipelines | Data and external connection configuration |
| Dataflows | Support depends on item generation/current tenant capability | Connections, credentials, and refresh schedules |
Shortcut cache rule used by the missing file exhibits
| Shortcut source/file | Cache result | Rule |
|---|---|---|
| Amazon S3, GCS, or S3-compatible; file ≤ 1 GB | Eligible for cache | Supported cross-cloud source and within per-file size limit. |
| Amazon S3, GCS, or S3-compatible; file > 1 GB | Read from source | Files over 1 GB are not cached. |
| ADLS Gen2 shortcut | Read from source | Shortcut caching does not apply to ADLS Gen2. |
Orchestration exhibit
| Requirement | Configuration |
|---|---|
| Notebook1 every weekday at 8:00 AM | A scheduled pipeline containing the Notebook activity. |
| Notebook2 when a Blob file is saved | A storage-event-triggered pipeline containing the Notebook activity. |
| Model1 after Notebook1 succeeds | An On Success dependency from Notebook1 to an explicit semantic-model refresh activity. |
5. Review the 20 supplied questions
Q1 · Give analysts access only to Gold Delta tables
Answer: C — share Lakehouse2 and grant Read all SQL endpoint data.
Do not place analysts in WorkspaceA. Spark/OneLake ReadAll would be broader than their SQL-only requirement, while Build permission is for report authoring rather than table access.
Do not place analysts in WorkspaceA. Spark/OneLake ReadAll would be broader than their SQL-only requirement, while Build permission is for report authoring rather than table access.
Q2 · Semi-structured data read by T-SQL, KQL, and Spark; written only by Spark
Answer: A — Lakehouse.
Spark writes open Delta data; Spark and the SQL endpoint can read it directly, and KQL can reach OneLake data through supported integration/shortcut patterns.
Spark writes open Delta data; Spark and the SQL endpoint can read it directly, and KQL can reach OneLake data through supported integration/shortcut patterns.
Q3–Q4 · Copy private on-premises SQL Server data to Warehouse
Answer: Data pipeline.
The two supplied questions test the same rule. Use Copy activity with a Fabric connection bound to the on-premises data gateway.
The two supplied questions test the same rule. Use Copy activity with a Fabric connection bound to the on-premises data gateway.
Q5 · Direct Lake performance degrades as a fact table grows
Answer: D — replace MD5 hash surrogate keys with an efficient numeric type such as BIGINT.
Integer relationship keys compress and join more efficiently than wide hash strings. This improves the model without immediately buying more capacity.
Integer relationship keys compress and join more efficiently than wide hash strings. This improves the model without immediately buying more capacity.
Q6 · Summarize order quantity by year/product and by year
Answer: GROUPING SETS.
Use the SQL shown above with groups for (year, product) and (year). ROLLUP would also produce a grand total unless you filter it.
Use the SQL shown above with groups for (year, product) and (year). ROLLUP would also produce a grand total unless you filter it.
Q7 · Choose three DimProduct columns
Answers: B, C, F — ProductName, ProductColor, ProductID.
Date, TransactionID, and SalesAmount describe the transaction/fact.
Date, TransactionID, and SalesAmount describe the transaction/fact.
Q8 · Attach two notebooks to the same Spark session
Answer: A — enable high concurrency for notebooks.
Dynamic allocation changes executor scaling, not session sharing.
Dynamic allocation changes executor scaling, not session sharing.
Q9 · Write Customer data without reading Employee PII
Answers: D, E, F.
Create Workspace2/Lakehouse2, move Employee there, and grant Contributor only in Workspace1. This isolates the restricted table because a workspace write role is broad.
Create Workspace2/Lakehouse2, move Employee there, and grant Contributor only in Workspace1. This isolates the restricted table because a workspace write role is broad.
Q10 · Warehouse object required for RLS
Answer: D — FUNCTION.
Create an inline table-valued predicate function with SCHEMABINDING, then attach it through a security policy.
Create an inline table-valued predicate function with SCHEMABINDING, then attach it through a security policy.
Q11 · What survives deployment?
Use the deployment matrix above.
The recurring traps are that data and schedules do not move with definitions, and target-stage bindings/credentials require separate configuration.
The recurring traps are that data and schedules do not move with definitions, and target-stage bindings/credentials require separate configuration.
Q12 · Add an Eventhouse to the Dev/Test/Prod promotion process
Answer: B — Fabric deployment pipeline.
Eventhouse is a supported Fabric deployment-pipeline item.
Eventhouse is a supported Fabric deployment-pipeline item.
Q13 · Detect invalid Warehouse references before deployment
Answer: A — SQL database project.
A project build validates the schema model and identifies unresolved references with less custom development than hand-written scripts.
A project build validates the schema model and identifies unresolved references with less custom development than hand-written scripts.
Q14 · Enable OneLake availability for an existing Eventhouse
Answer: E — only newly ingested data is exposed under the exam rule used by this guide.
Existing historical data is not automatically backfilled; enable OneLake availability early. If the live UI offers an explicit existing-data/backfill option, follow the delivered question’s wording.
Existing historical data is not automatically backfilled; enable OneLake availability early. If the live UI offers an explicit existing-data/backfill option, follow the delivered question’s wording.
Q15 · Authenticate an Azure DevOps pipeline calling Fabric APIs
Answer: A — service principal.
Use application authentication for unattended automation, not a stored user password.
Use application authentication for unattended automation, not a stored user password.
Q16 · Which GCS shortcuts use the cache?
Apply the recreated cache table.
The source screenshot with file sizes was absent: select every GCS-backed shortcut whose individual file is 1 GB or smaller. Do not guess a letter without the exhibit.
The source screenshot with file sizes was absent: select every GCS-backed shortcut whose individual file is 1 GB or smaller. Do not guess a letter without the exhibit.
Q17 · Spark job must access private Azure SQL
Answer: B — managed private endpoint.
This gives the Spark workload private connectivity to the Azure resource. The on-premises gateway pattern applies to pipeline connectivity and private on-premises sources.
This gives the Spark workload private connectivity to the Azure resource. The on-premises gateway pattern applies to pipeline connectivity and private on-premises sources.
Q18 · Cache results for ADLS and S3 shortcuts
Apply source plus file-size rules.
S3 files up to 1 GB are cache-eligible; files over 1 GB and ADLS Gen2 shortcuts are read from source. The missing source table prevents a responsible letter-only answer.
S3 files up to 1 GB are cache-eligible; files over 1 GB and ADLS Gen2 shortcuts are read from source. The missing source table prevents a responsible letter-only answer.
Q19 · Schedule, event trigger, and semantic-model dependency
Use the orchestration table above.
Notebook1 uses a weekday schedule, Notebook2 uses a Blob storage event trigger, and Model1 uses an explicit refresh after Notebook1 succeeds.
Notebook1 uses a weekday schedule, Notebook2 uses a Blob storage event trigger, and Model1 uses an explicit refresh after Notebook1 succeeds.
Q20 · Create a domain, subdomains, and assign workspaces
Answer: D — Fabric administrator.
Domain Admin can manage an existing assigned domain, but creating a top-level domain requires Fabric administrator authority.
Domain Admin can manage an existing assigned domain, but creating a top-level domain requires Fabric administrator authority.
🔑 Complete quick reference — all key comparisons from every module
✅ Reviewed question bank — 176 deduplicated questionsChoices are concise and parallel so answer length does not reveal the key. The full reasoning appears only after you select an answer. Answers and explanations were checked against the current Microsoft DP-700 study guide and Microsoft Learn documentation.
Q 1 of 32Correct: 0 · Answered: 0
0%
Score on questions
← Previous → Next 1–4 Answer