Orchestrating Multi-Region EFS Data Processing with Step Functions

Why the Answer Is Correct

Option C is the correct answer because it aligns with the principle of 'moving compute to data' rather than moving data to compute, which minimizes network transfer costs and latency. By deploying Lambda functions to each region, the engineer avoids cross-region data movement entirely. The output is then stored in a central S3 bucket, which is a standard, low-effort aggregation pattern. This approach respects the requirement that the data science team's environment remains in a single region while allowing the processing to happen locally where the EFS files reside.

Why the Other Options Are Wrong

Option A involves VPC peering across multiple regions and mounting remote EFS file systems into Lambda. While technically possible with some configuration, it introduces significant networking complexity and potential latency issues, violating the 'least effort' constraint. Option B suggests replicating EFS data, but EFS does not natively support active-active replication to another region in a simple manner without additional services like DataSync or third-party tools, making it high effort. Option D uses AWS DataSync to consolidate all data into one region. This violates the spirit of efficiency by moving large volumes of data across regions just for processing, whereas Option C processes locally and moves only small output files to S3.

Community Comment Notes

The community split between C and D, with many favoring D due to a misunderstanding of 'effort'. As user HagarTheHorrible noted, 'Data Sync is for large scale migration, Lambdas would do just fine here... C', highlighting that DataSync is often overkill for this specific orchestration scenario. Another comment simply stated 'Peer the VPC introduce complexity, D is a much better solution', reflecting the common trap of preferring centralized data over distributed compute.