All 33 lessons
Start here: cloud and data basics
- 1. Start here: the exam, the plan and what you already know
- 2. Cloud in plain words: SaaS, PaaS, IaaS and where Fabric fits
- 3. Data basics: files, tables, formats and the lakehouse idea
- 4. A tour of Microsoft Fabric: OneLake, workspaces and items
- 5. The three languages: SQL, PySpark and KQL side by side
- 6. Getting data in: the map of Fabric tools
Store, ingest and transform batch data (exam area 2, 30-35%)
- 7. Choosing a data store and a transformation tool
- 8. Shortcuts and mirroring: use data without copying it
- 9. Ingest data with pipelines and Copy job
- 10. Transform with PySpark and SQL: join, denormalize, aggregate
- 11. Messy data: duplicates, missing values and late arrivals
- 12. Full and incremental loads, and loading a star schema
Orchestrate and stream (exam areas 1 and 2)
Set up and secure a Fabric solution (exam area 1, 30-35%)
- 18. Workspace settings: Spark, domains, OneLake and data workflows
- 19. Version control with Git and database projects
- 20. Deployment pipelines: development, test and production
- 21. Who can see what: workspace roles and item permissions
- 22. Fine-grained security: rows, columns, objects, files and masking
- 23. Governance: sensitivity labels and endorsement
Monitor, fix and optimize (exam area 3, 30-35%)
- 24. Monitoring hub, alerts and Activator
- 25. Fixing errors in pipelines, dataflows and notebooks
- 26. Fixing errors in Eventhouse, Eventstream and T-SQL
- 27. Optimize lakehouse tables: V-Order, OPTIMIZE, VACUUM and partitions
- 28. Optimize Spark, pipelines and warehouses
- 29. Optimize eventstreams, eventhouses and queries
Exam practice and final review
20 min
Final review: every decision on one page
Review every 'choose between' decision and key fact the DP-700 exam tests, then plan the night before and the morning of.
By the end of this lesson you can
- Pick the right store, tool or setting for any common DP-700 scenario from one set of tables
- Recall the roles, permissions, window types, maintenance commands and limits the exam names
- Run a final 20-card review and a few last scenario checks
- Plan the night before and the morning of the exam
From earlier lessons · 1 of 3
You have 35 minutes left and the case study hasn't started. What do you do first?
Pick one answer.
From earlier lessons · 2 of 3
In a yes/no series, two different proposed solutions both seem to meet the goal. How do you answer?
Pick one answer.
From earlier lessons · 3 of 3
A scenario ends: "...with the least development effort." Two options work: a notebook and a Copy job. Which fits?
Pick one answer.
Over 32 lessons you built Northwind Rides' whole platform: scooter events streaming into an Eventhouse, a mirrored payments database, nightly pipelines, secured Lakehouse tables and tuned queries. Tomorrow you sit DP-700. You don't need anything new tonight. You need every decision you made, on one page.
Where data lives (lessons 3, 4 and 7)
Exam depth: know it cold.
| If the scenario says | Choose |
|---|---|
| Files and Delta tables, Spark or PySpark skills, big data or ML | Lakehouse. |
| T-SQL team, star schema, multi-table transactions | Warehouse. |
| Streaming events, time series, KQL | Eventhouse. |
| An app database with many small reads and writes (OLTP) | SQL database in Fabric. |
| T-SQL users only read Lakehouse tables | the Lakehouse's SQL analytics endpoint (read-only). |
Rule: pick by the team's language and the need for transactions. All of them keep data in OneLake. Eventhouse data reaches OneLake as Delta only when you turn on OneLake availability.
Getting data in (lessons 6, 8 and 9)
Exam depth: know it cold.
| If the scenario says | Choose |
|---|---|
| Use data in place, no copy, from ADLS, S3 or another workspace | Shortcut. |
| Replicate a database continuously, free, little setup | Database mirroring (a read-only Delta copy in OneLake). |
| Use an Azure Databricks catalog's tables without copying | Metadata mirroring (creates shortcuts, copies no data). |
| Copy data in bulk or incrementally, no pipeline to build | Copy job. |
| Copy as one step among others, with parameters and custom queries | Copy activity in a pipeline. |
| Live events with low latency | Eventstream. |
Rule: keep it in place, then replicate, then copy. Each step adds a copy to manage. Only a Copy job has built-in watermark-based incremental copy.
Transforming data (lessons 5, 7, 10 and 11)
Exam depth: know it cold.
| If the scenario says | Choose |
|---|---|
| No code, Power Query skills, a business analyst maintains it | Dataflow Gen2. |
| Complex logic, huge volumes, Python or Spark libraries | Notebook (PySpark). |
| Data already in a Warehouse, SQL team | T-SQL (views, stored procedures). |
| Events in an Eventhouse, time-based analysis | KQL. |
Rule: match the tool to the people who maintain it, then to the data's size and location.
Loading patterns (lessons 11 and 12)
Exam depth: know it cold.
| If the scenario says | Choose |
|---|---|
| A small table, or a source with no change tracking | Full load (replace everything). |
| A large table with a modified-date or ID column | Incremental load with a watermark. |
| Overwrite the old value, history not needed | Slowly changing dimension Type 1. |
| Keep every past value with dates | Slowly changing dimension Type 2 (new row per change). |
| Duplicate rows from retries | Deduplicate on the business key, keep the latest. |
| Late events in a Spark stream | A watermark: how late an event may arrive and still count. |
A watermark in batch loads is the last value you loaded. In streaming it's a lateness limit. Same word, two meanings.
Orchestrating (lessons 13, 14 and 15)
Exam depth: know it cold.
| If the scenario says | Choose |
|---|---|
| Run several steps in order, low code, with retries and branches | Pipeline. |
| The logic itself is code, run interactively or as one step | Notebook (called from a pipeline if needed). |
| Run at fixed times | Schedule. |
| Run when a file lands in OneLake or Blob Storage | Event trigger (storage event). |
| Same pipeline, different table each run | Parameters plus dynamic expressions. |
| Use the new file's name in the run | @pipeline()?.TriggerEvent?.FileName. |
Rule: the pipeline conducts, the notebook plays. Event triggers run on Activator behind the scenes.
Streaming and windows (lessons 16 and 17)
Exam depth: know it cold.
| If the scenario says | Choose |
|---|---|
| No-code ingest, filter and route events | Eventstream. |
| Spark code, Delta Lake output, complex stateful logic | Spark Structured Streaming. |
| Query and analyze events already stored | KQL in an Eventhouse. |
Window types:
| Window | What it does |
|---|---|
| Tumbling | fixed size, no overlap; each event in one window. |
| Hopping | fixed size, overlaps; hop smaller than size. |
| Sliding | outputs only when an event enters or leaves. |
| Session | groups events close together; closes after a gap. |
| Snapshot | groups events with the same timestamp. |
A tumbling window is a hopping window whose hop equals its size. In KQL, summarize ... by bin(Time, 5m) puts each row in exactly one fixed 5-minute bucket: the same result as a tumbling window.
Settings and lifecycle (lessons 18, 19 and 20)
Exam depth: know it cold.
| If the scenario says | Choose |
|---|---|
| Spark sessions must start in seconds | Starter pool (Medium nodes, prewarmed). |
| Several notebooks should share one session | High concurrency mode. |
| Group workspaces by business area | Domain. |
| Track item changes and branch | Git integration: Azure DevOps, GitHub or GitHub Enterprise. |
| Promote content from development to test to production | Deployment pipeline (2 to 10 stages). |
| Manage Warehouse schema as code | Database project (builds a .dacpac schema file). |
Only a workspace Admin connects a workspace to Git. A deployment pipeline's stage count can't change after you create it.
Security and governance (lessons 21, 22 and 23)
Exam depth: know it cold.
| If the scenario says | Choose |
|---|---|
| Access to one item only | Share the item (item permissions). |
| Read everything in the workspace, change nothing | Viewer. |
| Create and edit items, but not add people | Contributor. |
| Add people and share, but not delete the workspace | Member. |
| Hide some rows per user | Row-level security (RLS). |
| Hide some columns | Column-level security (CLS). |
| Hide whole tables or views | Object-level security (OLS). |
| Show only part of an email address | Dynamic data masking (DDM). |
| Grant read on some Lakehouse folders | OneLake security role. |
| Mark an item as confidential | Sensitivity label. |
| Mark an item as trustworthy | Endorsement: Promoted or Certified. |
| Mark a data item as the single source of truth (customer list, product codes) | Endorsement: Master data. |
Rule: grant the smallest scope that does the job. DDM alone doesn't stop users who can query from guessing values, so pair it with CLS or RLS.
Monitoring and fixing (lessons 24, 25 and 26)
Exam depth: know it cold.
| If the scenario says | Choose |
|---|---|
| See every run and its status in one place | Monitor hub. |
| Email or Teams alert, or start an item, when a value crosses a line | Activator rule (Fabric's rules-and-alerts engine). |
| Email when a scheduled pipeline fails | Failure notifications on the schedule. |
| Find a broken step in an Eventstream | the Authoring errors tab in Edit mode. |
| Find slow or failed Warehouse queries | queryinsights views (built-in query history). |
| Warehouse query slow only on its first run | Cold start: data loads into cache first; judge later runs. |
Rule: find where it failed (Monitor hub), read the error, then fix the step, not the whole flow.
Optimizing (lessons 27, 28 and 29)
Exam depth: know it cold.
| Symptom | Fix |
|---|---|
| Many small Parquet files, slow reads | OPTIMIZE. |
| Storage grows with old unreferenced files | VACUUM. |
| Table read many times by Power BI or the Warehouse | V-Order on. |
| Table written once and read once | V-Order off. |
| Heavy Spark SQL on Parquet or Delta | Native execution engine. |
| Small table joined to a huge one in Spark | Broadcast join. |
| Slow Copy activity on a big source | Degree of copy parallelism and partitioned reads. |
| Same Warehouse query repeated | Result set caching (on by default). |
| Bad Warehouse query plans | Up-to-date statistics, matching data types. |
| Slow KQL on recent data | Hot cache covers it, time filter first, has. |
| Repeated KQL summary on a dashboard | Materialized view. |
| Too much ingested from an Eventstream | Filter before the destination. |
Facts to know cold
- Only Admin can delete the workspace, add admins or connect Git. Member adds people with lower roles. Contributor creates, edits and runs items. Viewer only reads.
- Viewer reads Lakehouse and Warehouse data with T-SQL (ReadData), not through Spark or OneLake APIs (ReadAll).
- OneLake security roles grant Read or ReadWrite on folders and tables. They only limit Viewers and Read users; Admin, Member and Contributor already see all data.
- Five window types: tumbling, hopping, sliding, session, snapshot. Hopping and sliding windows can overlap, so one event can count in several.
- VACUUM keeps 7 days by default. Fabric rejects shorter retention unless you turn off a safety check.
- V-Order costs about 15% more write time and can give up to 50% more compression.
- Eventhouse caching and retention default to 3,650 days. Caching must be lower than or equal to retention.
- Eventstream keeps events 1 day by default, up to 90. Throughput: Low under 10 MB/s, Medium 10 to 100, High over 100.
- Deployment pipelines have 2 to 10 stages, 3 by default. Deploying overwrites the paired items in the next stage.
- Starter pools usually start a session in 5 to 10 seconds. High concurrency shares one session among up to 5 notebooks by default.
- Promoted: anyone with write permission on the item. Certified and Master data: only users the Fabric admin names. Master data applies only to items that hold data.
Exam-style practice
You are a data engineer at Northwind Rides. Payments live in Azure SQL Database. Finance wants them in Fabric for reports, kept in sync, with minimal setup and no extra cost for the replication itself. What should you use?
Pick one answer.
You are a data engineer at Northwind Rides. Contractors must read only the Files/maintenance folder of the ops Lakehouse, with no write access and nothing else in the workspace. What should you configure?
Pick one answer.
You are a data engineer at Northwind Rides. You need the count of trips per station for every 5-minute period. Periods must not overlap, and each trip counts once. Which window should the Eventstream Group by use?
Pick one answer.
You are a data engineer at Northwind Rides. A Lakehouse table gets small writes all day. Power BI reads it in Direct Lake many times an hour, and reads keep slowing down. You need faster reads. What should you run?
Pick one answer.
The night before and the morning of
The night before
- Read this page's tables and facts once. Run the 20 cards once. Then stop.
- Check your ID: original, unexpired, government-issued, with a name that matches your certification profile exactly.
- Testing online? Run the system test on the same computer, in the same room. Clear the desk, unplug extra monitors, and plan a room with a closed door.
- Sleep. A rested brain reads constraint words better than a tired one that crammed.
The morning of
- Eat, and start check-in early. Online, you photograph your ID and scan the room before a proctor is assigned.
- Keep your phone out of reach but where you can hear it, in case the proctor calls.
- In the exam: last sentence first, constraint word, cross out, choose. Keep 20 to 30 minutes for the case study.
- Answer every question. There's no penalty for a guess.
Worth keeping
- Pick stores and tools by the team's language, the latency needed, and whether data may be copied.
- Grant the smallest scope that works: item before workspace, folder before item, row or column before table.
- Optimize by engine: OPTIMIZE and V-Order for Delta, caching and statistics for the Warehouse, hot cache and has for KQL.
- Know the numbers: 7-day VACUUM, 3,650-day Eventhouse defaults, 2 to 10 deployment stages, 1 to 90 days Eventstream retention.
- Tonight: review, check ID and setup, sleep. Tomorrow: find the constraint word and answer everything.
On the job
These tables don't expire after the exam. At work, a new request arrives as one sentence. "Finance wants live payments in Power BI, but IT won't allow another copy." Find the constraint, look down the left column, and you have a first design in a minute. Then prove it on a trial workspace.
Keep this page open for your first few projects. Good luck tomorrow, from the whole team.
Leaderboard
Loading…
50 XP for each lesson you finish, 10 for each right answer. Names show only for learners who turn on Show me on leaderboards.
Sources for this lesson (47)
Each fact in the lesson is marked with the number of the page it comes from.
- [4]Microsoft Learn: Study Guide for Exam DP-700
- [7]Microsoft Learn: Exam scoring and score reports
- [48]Microsoft Learn: Overview of Microsoft Fabric eventstreams
- [65]Microsoft Learn: bin()
- [67]Microsoft Learn: Get data into Microsoft Fabric
- [68]Microsoft Learn: Microsoft Fabric decision guide: copy activity, Copy job, dataflow, Eventstream, or Spark
- [69]Microsoft Learn: Microsoft Fabric decision guide: Choose a data movement strategy
- [70]Microsoft Learn: Microsoft Fabric decision guide: Choose a data integration strategy
- [73]Microsoft Learn: Microsoft Fabric decision guide: choose a data store
- [102]Microsoft Learn: Stateful processing with Structured Streaming
- [114]Microsoft Learn: Run, schedule, or use events to trigger a pipeline
- [133]Microsoft Learn: Process event data by using the event processing editor
- [137]Microsoft Learn: Overview of Structured Streaming
- [146]Microsoft Learn: Windowing functions (Azure Stream Analytics and Fabric Eventstream)
- [150]Microsoft Learn: Materialized views
- [155]Microsoft Learn: Overview of Apache Spark compute in Microsoft Fabric
- [160]Microsoft Learn: High concurrency mode in Apache Spark for Fabric
- [168]Microsoft Learn: What is Microsoft Fabric Git integration?
- [182]Microsoft Learn: Develop warehouse projects in Visual Studio Code
- [184]Microsoft Learn: Introduction to deployment pipelines
- [185]Microsoft Learn: The deployment pipelines process
- [191]Microsoft Learn: Roles in workspaces in Microsoft Fabric
- [196]Microsoft Learn: Data security in OneLake
- [200]Microsoft Learn: Dynamic data masking in Fabric Data Warehouse
- [202]Microsoft Learn: Secure your Fabric Data Warehouse
- [209]Microsoft Learn: Apply sensitivity labels to Fabric items
- [213]Microsoft Learn: Endorsement overview
- [237]Microsoft Learn: Dataflow Gen2 data destinations and managed settings
- [245]Microsoft Learn: Permission model
- [262]Microsoft Learn: Spark monitoring and performance optimization best practices
- [264]Microsoft Learn: Performance guidelines in Fabric Data Warehouse
- [266]Microsoft Learn: VACUUM Delta tables
- [267]Microsoft Learn: Run Delta table maintenance in lakehouse
- [273]Microsoft Learn: Native execution engine for Fabric Data Engineering
- [276]Microsoft Learn: Copy activity performance and scalability guide
- [279]Microsoft Learn: Result set caching
- [280]Microsoft Learn: Change data policies
- [281]Microsoft Learn: Caching policy (hot and cold cache)
- [283]Microsoft Learn: String operators
- [286]Microsoft Learn: Configure settings for a Fabric eventstream
- [289]Microsoft Learn: About online exams with Pearson VUE
- [290]Microsoft Learn: Exam and assessment lab security policy
- [291]Microsoft Learn: Frequently asked questions (FAQs) about online proctored exams
- [293]Microsoft Learn: Best practices for OneLake security
- [296]Microsoft Learn: Fabric Activator rules
- [297]Microsoft Learn: Hopping Window
- [298]Microsoft Learn: Governance overview and guidance