Reviewed by Aditya Kumar · Last reviewed 2026-03-24
**Why merge small files**: Small files (e.g., <128 MB) hurt Athena/Spark—each file incurs metadata overhead, and many small files cause task explosion. Athena charges per MB scanned; small-file overhead increases effective cost. Target 128–256 MB per file for optimal scan...
This hard-level Cloud/Tools question appears frequently in data engineering interviews at companies like Capco. While less common, it tests deeper understanding that distinguishes strong candidates. Mastering the underlying concepts (etl, partition, spark) will help you answer variations of this question confidently.
This is a senior-level question that tests architectural thinking. Lead with the high-level design, then drill into specifics. Discuss trade-offs explicitly - there is rarely one correct answer. Show awareness of scale, fault tolerance, and operational complexity.
Why merge small files: Small files (e.g., <128 MB) hurt Athena/Spark—each file incurs metadata overhead, and many small files cause task explosion. Athena charges per MB scanned; small-file overhead increases effective cost. Target 128–256 MB per file for optimal scan throughput. Architecture: Glue ETL job reads source, applies coalesce(N) or repartition(N) based on target file count, writes Parquet/ORC. For incremental: use Glue bookmarks so only new files are processed—run compaction separately to avoid blocking ingest. Scalability trade-off: Coalesce reduces parallelism; if input is 10K files, coalesce(4) runs 4 tasks—balance with job duration. Run compaction during off-peak. Cost: Glue bills DPU-minute; coalesce reduces task count but may increase per-task time—profile. A daily compaction job for 100 GB typically runs 15–30 min at 10 DPUs. Best practice: Separate ingest pipeline from compaction; do not block real-time writes. Use a lifecycle: raw (many small) → compaction job → curated (larger files).
Pro-Move: 'We had 2M small files from Kinesis Firehose causing 5× Athena cost—added a nightly Glue compaction to 50K files, cutting scan time 70%.' Red Flag: Merging in the same job as ingest without bookmarks—risk of reprocessing or blocking.
Some links below are affiliate links. If you buy through them we may earn a small commission at no extra cost to you — it helps keep DataEngPrep free.
According to DataEngPrep.tech, this is one of the most frequently asked Cloud/Tools interview questions, reported at 1 company. DataEngPrep.tech maintains an editor-reviewed database of 1,863 data engineering interview questions across 7 categories.