Data lake design patterns in Hadoop - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Understanding the Zone Architecture in Data Lakes

In a typical data lake design, data is organized into different zones such as raw, cleansed, and curated. What is the primary purpose of the raw zone?

A Store metadata and data catalog information.

B Store only cleaned and validated data ready for business use.

C Store aggregated and summarized data for reporting.

D Store data exactly as ingested without any transformation or filtering.

💡 Hint

Think about where data first lands in the data lake before any processing.

The raw zone holds data exactly as it arrives from source systems. It is unprocessed and serves as the foundation for further cleansing and transformation.

❓ data_output

intermediate

Data Partitioning Effect on Query Performance

Given a Hadoop data lake storing logs partitioned by date, what is the expected effect on query performance when filtering by a specific date?

A Query performance is unaffected by partitioning.

B Query performance worsens because all partitions must be scanned.

C Query performance improves because only the relevant partition is scanned.

D Query performance depends only on file size, not partitioning.

💡 Hint

Consider how partition pruning works in Hadoop query engines.

Partitioning data by date allows query engines to scan only the relevant partition, reducing data scanned and improving performance.

🔧 Debug

advanced

Identifying the Cause of Data Duplication in a Data Lake

A data lake ingestion pipeline writes data daily to the curated zone. After some time, duplicate records appear in the curated data. Which of the following is the most likely cause?

A The ingestion job appends data without checking for existing records.

B The raw zone data is corrupted causing duplicates.

C Partitioning by date was not applied in the raw zone.

D The data lake storage is running out of space.

💡 Hint

Think about how data is merged or appended during ingestion.

If the ingestion job appends data daily without deduplication or overwrite logic, duplicates accumulate in the curated zone.

🚀 Application

advanced

Choosing the Best Data Lake Design Pattern for Streaming Data

You need to design a data lake solution for real-time streaming data ingestion and analytics. Which design pattern is most suitable?

A Using a traditional data warehouse instead of a data lake.

B Lambda architecture combining batch and speed layers.

C Storing all data in the raw zone without transformation.

D Only batch processing with nightly ingestion jobs.

💡 Hint

Consider architectures that handle both real-time and batch data.

Lambda architecture supports both real-time streaming (speed layer) and batch processing (batch layer), making it ideal for streaming data analytics.

❓ Predict Output

expert

Output of Hive Query on Partitioned Data

Consider a Hive table partitioned by year and month. What is the output of the following query?

SELECT year, month, COUNT(*) AS cnt FROM sales WHERE year = 2023 AND month = 5 GROUP BY year, month;

Assume the table has 1000 records for May 2023 and 5000 records for other months.

A [{'year': 2023, 'month': 5, 'cnt': 1000}]

B [{'year': 2023, 'month': 5, 'cnt': 6000}]

C [{'year': 2023, 'month': 5, 'cnt': 5000}]

D Empty result set

💡 Hint

Focus on the filter conditions and grouping in the query.

The query filters data for year 2023 and month 5, then counts records. Since there are 1000 records for May 2023, the count is 1000.