The Volume-Doubling Question

What the interviewer is checking

Underneath the question, the interviewer wants to know whether you locate the bottleneck before proposing a solution. Saying 'add more consumers' without first ruling out ingestion and sink writes as the bottleneck tells them you scale by instinct. The FAANG system design rubric scores 'trade-offs and bottlenecks' at 20-25% of the total grade. This is the section where L5 candidates earn their hire, and where L6 candidates earn their strong hire.

Find the slowest link first

Every pipeline has a throughput chain: source rate, ingestion layer, processing layer, sink write speed. The slowest link sets system throughput, so scaling any other link costs money and changes nothing. Find that binding constraint before you add hardware.

The 60-second framework

• Quantify: 'At 1B events/day, that's ~11,500 events/second sustained, with 3-5x spikes during peak.'
• Identify the chain: 'Source rate > Kafka ingestion > consumer processing > warehouse sink.'
• Find the bottleneck: 'If each consumer processes 2,000 events/sec, I need 6 consumers for sustained load and 18 for peak.'
• State the scaling mechanism: 'Add Kafka partitions and consumers together. One consumer per partition.'

Step 1 is the strong-hire signal. Most candidates never convert 'lots of data' into events/second, and the interviewer is waiting for someone to do that math and state the actual throughput requirement.

About this lesson

This section is part of the Backpressure and Scaling: Advanced lesson on DataDriven, the home of data engineering. Lessons cover SQL, Python, Spark, data modeling and pipeline architecture, and each section pairs its explanation with worked examples and practice you run as you go.