← Google Interview Insights

Google·Software Engineer·Technical Phone Screen·Senior

Senior
Jun 2026

Summary

Prepped for a Google data engineering screen and got a classic big data trade-off question. Not much to report beyond that one question, but it's the kind of thing that trips people up if they haven't thought it through carefully.

Questions Asked (1)

Q1

In what scenarios would you prefer Hadoop over PySpark for data processing?

Technical Trade-offsSystem Design
Author's notes

I knew the broad strokes but fumbled the specifics a bit.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Acknowledge that PySpark is generally preferred for its speed, ease of use, and rich API, but identify specific scenarios where Hadoop MapReduce still has advantages. Focus on trade-offs like batch processing of massive datasets, existing Hadoop investments, and operational simplicity in certain contexts.

Pro tip: Emphasize that the choice often depends on the existing ecosystem and team expertise, not just technical merits. Showing awareness of total cost of ownership and migration efforts demonstrates maturity.

1. Clarify the context

Start by stating that the decision depends on factors like data size, latency requirements, existing infrastructure, and team skills. This shows you consider the broader picture.

2. Highlight Hadoop's strengths

Discuss scenarios where Hadoop MapReduce excels: extremely large-scale batch processing, environments with existing Hadoop clusters, and cases where disk-based processing is acceptable.

3. Compare with PySpark

Contrast with PySpark's advantages: in-memory computation, interactive analysis, and support for iterative algorithms. Explain why PySpark is usually preferred for these.

4. Consider operational factors

Mention non-technical aspects like cost, maintenance, and team familiarity. For example, if a company already has a Hadoop ecosystem, using MapReduce might be simpler.

5. Conclude with a balanced view

Summarize that while PySpark is often the default, Hadoop remains relevant in specific legacy or large-scale batch scenarios. Avoid absolute statements.

Key Points to Mention

  • Batch processing of petabyte-scale data where disk-based processing is sufficient
  • Existing Hadoop infrastructure and investment (e.g., HDFS, YARN, Hive)
  • Scenarios requiring strict data locality and fault tolerance without in-memory overhead
  • Team expertise and operational simplicity in Hadoop ecosystem
  • Cost considerations: Hadoop may be cheaper for long-running batch jobs on existing clusters
  • PySpark's advantages: speed, ease of development, interactive analysis, and iterative algorithms

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.