System Architecture & Design Flashcards

In a medallion lakehouse architecture, which layer typically holds cleansed, conformed, and deduplicated data that is not yet aggregated for business reporting?

Answer: Silver

Bronze holds raw data, silver holds cleansed and conformed data, and gold holds business-level aggregates.

A data lake applies structure only when data is queried, not when it is loaded. What is this approach called?

Answer: Schema-on-read

Schema-on-read stores raw data and interprets its structure at query time, which gives flexibility at ingestion.

Which slowly changing dimension type keeps full history by adding a new row with effective dates each time an attribute changes?

Answer: Type 2

SCD Type 2 adds a new versioned row for every change and keeps all earlier rows for historical analysis.

Why are columnar file formats such as Parquet preferred for analytical workloads?

Answer: They allow column pruning and high compression for scan-heavy queries

Columnar storage reads only the needed columns and compresses similar values well, which speeds up analytics.

How does a snowflake schema differ from a star schema?

Answer: Its dimension tables are normalized into multiple related tables

A snowflake schema normalizes dimensions into sub-dimensions, which reduces redundancy but adds joins.

Which architecture processes all data, both real-time and historical, through a single stream-processing path and reprocesses history by replaying the log?

Answer: Kappa architecture

Kappa removes Lambda's separate batch layer and treats everything as a replayable stream.

A table of billions of event records is mostly queried by date range. Which physical design choice improves query performance the most?

Answer: Partitioning the table by event date

Partitioning by date lets the engine skip partitions outside the queried range (partition pruning).