Objective and core distinction
Batch processing works on a bounded collected dataset. Stream processing handles an ongoing arrival sequence. Both can filter or aggregate data, but their input and timing assumptions differ. A complete pipeline must also distinguish ingestion or buffering from processing. Capturing a message does not itself calculate a running total. Likewise, calling a solution real-time does not prove zero delay or correct failure handling. Write what arrives, when the input collection is considered complete, what computation is needed, and when an output must be available. Those conditions can justify a pattern more precisely than a service slogan.
Worked example
Fictional case:a book festival counts arrivals during its opening day and also creates a reconciled end-of-day attendance report. The running display needs results updated as events arrive; the report needs a bounded collection checked after the day closes. A stream-processing path can address the first requirement while a batch task addresses the second. The two outputs are not interchangeable. Saving incoming events in a buffer supports later processing, but the display still needs a consumer that applies the counting logic and publishes a result. Preserve the original events if reconciliation must later inspect discrepancies. The conclusion concerns output requirements, not a universal preference for streams.
Changed condition and misconception
In a supplied counting exercise, a window can be defined with an inclusive start and exclusive end. Apply that declared rule to every event rather than assume a vendor default. A timestamp exactly at the upper bound then differs from one inside the interval. Actual systems have more timing and recovery considerations than this toy rule specifies; this lesson does not teach watermark configuration or promise every late event is handled in one way. Fabric Real-Time hub supplies a central discovery role for supported streaming sources. It is distinct from a finished batch table or a semantic-model measure. Source discovery, buffering, processing and reporting are related but separate stages.
Transfer check
Transfer check:a fictional team receives sensor messages immediately but runs its only calculation each night. Someone calls the displayed result streaming because input capture is continuous. What fact should the explanation separate? Check basis:arrival/capture and computation are different operations. The output is still tied to the nightly calculation unless another processing path exists. Then consider a second requirement for a fixed weekly reconciliation:waiting for that bounded set may be appropriate. State the output deadline and collection boundary before selecting a pattern. Use the linked tasks to distinguish pattern choice, declared-window arithmetic, missing computation and streaming-source discovery.
Try it, then explain it.
Complete a short set without looking at the example. After each answer, explain which evidence or rule supports your decision and why the alternatives do not fit.
Sources & authorship
Original OpenExam explanations and examples. References support the concepts; they are not the source of copied questions. Independent educator review is pending.
What is Microsoft Fabric - Microsoft Fabric | Microsoft LearnBig Data Architectures - Azure Architecture Center | Microsoft Learn