DP-900 Practice Worksheets & Answer Key

Answer: JSON device records whose optional property names vary from device to device; Tagged event records whose permitted payload fields vary by event type

1. Classify the representation described, rather than deciding from the product name alone. The varying optional properties and event-specific payload fields both provide organization without one fixed set of fields for every record.
2. The declared CSV columns and SQL columns define fixed tabular schemas in these descriptions. Audio recordings are the unstructured media example. JSON can be used with a fixed schema too; the stated field variation controls this case.
3. Check a changed condition: if the event contract instead required exactly the same declared fields on every record, variable payloads would no longer be the stated reason to call it semi-structured. The answer is tied to the descriptions, not to a permanent ranking of file extensions. Audio metadata can be tabular without making the audio itself a fixed-schema record.

Answer: 3 rows

1. Enumerate matching pairs under the equality condition. C1 matches S1 and S2, giving two output rows; C2 matches S3, giving one. S4 has no matching club and contributes none.
2. The result has 2 + 1 = 3 rows. Counting clubs gives 2, retaining every session gives 4, and dropping the join condition gives a 2 × 4 Cartesian product of 8.
3. Enumerate pairs rather than source-table sizes: C1 matches S1 and S2, C2 matches S3, and the intentionally unmatched S4/C3 has no club partner. Three matching pairs remain. The absence of a rejecting foreign-key constraint permits S4 in the source but does not make it join successfully. A validity rule on stored data and a condition selecting result pairs are different mechanisms.

Answer: Use Azure Files over SMB for application A; Use Azure Blob Storage for application B’s existing Blob REST requests

1. Keep the two interfaces separate: A needs SMB file sharing; B needs the Blob REST interface for object operations. Azure Files addresses the former, while Blob Storage addresses the latter.
2. A file-like capability is not automatically the same wire protocol. Blob Storage supports NFS in applicable configurations, but that does not turn NFS into the required SMB endpoint. Likewise, an Azure Files REST endpoint is not the unchanged Blob REST interface.
3. Test each application's actual endpoint rather than choosing one service for the broad word files. SMB is the first unchanged interface; Blob REST is the second. A supported alternative protocol does not automatically accept the requests already sent by either client. If an application rewrite were allowed, the candidates might change, but the present comparison intentionally fixes the client contracts rather than comparing all storage uses.

Answer: Store the original files in a data lake and parse the required fields in later analysis steps

1. The contract defers analysis-specific parsing and keeps the primary source collection file-oriented. A lake of original files directly provides that organization.
2. The other plans can preserve information, but make a fixed relational, dimensional or operational structure the primary access path before the stated later analyses. This does not claim warehouses can never retain raw data; it compares these supplied processing contracts.
3. The primary access contract is original files with analysis-specific parsing deferred. A fixed dimensional model could be useful later, but making it the initial required structure imposes a decision the stem postpones. The alternatives are not rejected for destroying all information; they are rejected for changing when and how readers must interpret it. A sound explanation should compare that contract rather than claim warehouses can never hold source data.

Answer: Send B’s replacement with If-Match E1; treat a precondition failure as a reason to reread and reconcile

1. The comparison must involve the current server item. B's If-Match E1 condition no longer matches E2, so the conditional write is rejected with a precondition failure rather than silently replacing the newer version. The client should fetch the current item and decide how to reconcile the business change.
2. Merely obtaining the new tag and applying an old calculation defeats the intended lost-update protection. Likewise, an unconditional successful write only says that a write was accepted; it does not show that B calculated from the latest state. The stem limits this case to the described single-region workflow rather than generalizing every multi-region conflict policy.

Answer: Treat columnar layout as relevant to selective column reads, then measure the actual scan and update workloads rather than infer a universal winner

1. A columnar representation can be relevant when a reader needs a few columns, but the format fact is not a benchmark for all operations. The two workloads touch data differently. Specify representative queries, required updates, correctness and resource conditions before comparing actual behavior.
2. One favorable scan cannot establish every-query superiority, just as one expensive update cannot establish that the format is unsuitable for every analytic scan. The comparison needs the access pattern and measured outcomes. This question makes no Azure cost or latency claim and does not convert a layout characteristic into a guarantee.

Answer: 3 records

1. Separate the header from the data lines. C1, C2 and C3 are the three records; the zero Units value does not remove C3.
2. Counting the header gives4. Adding Units gives3+7+0=10 units rather than a record count. Counting the two column names gives2 columns.
3. Mark the header as schema information, then enumerate C1, C2 and C3 as the three records. C3's zero Units is still a supplied client record; omitting it changes the population from clients to clients with positive units. Counting four would instead mix one header line with three data records. This is an explicit simple export model, not a claim that every physical CSV line is a complete record.

Answer: Azure SQL Database; Azure SQL Managed Instance

1. SQL Database and Managed Instance provide managed SQL Server database platforms. Their common maintenance boundary does not mean their supported feature scopes are identical.
2. Running database software on a self-administered VM or physical host is a different operating model. Adding the word Azure to that host does not make it one of these two managed SQL Server PaaS offerings.
3. Keep the responsibility boundary explicit: delegating underlying maintenance does not delegate every data definition, access decision, query design or business recovery requirement. The two named PaaS offerings satisfy the question's managed-platform description, while the customer-administered guest VM does not. A service supporting the SQL Server engine can belong to a different operating model; matching the engine name alone does not settle who maintains its host.