Interview Prep Hub

Data Systems Design

Prep for the 60-minute, discussion-based data systems design interview — a whiteboard-friendly conversation about architecting scalable, reliable data platforms. It turns on the topics the interviewer names directly: data ingestion pipelines, change data capture (CDC) strategies, and the big architectural trade-offs — batch vs streaming, data lakes vs warehouses vs lakehouses, and build vs buy. This guide gives you a repeatable design method, the building blocks, and a defensible point of view on every fork the conversation can take.

How to read this guide — the method

A systems-design discussion is graded on structured breadth and defended trade-offs, not on reciting one blessed architecture. The whole guide hangs off a repeatable six-step method you can run on any scenario — and the trade-off chapters give you the decision matrices that actually win the conversation.

Step	What you do
1 · Requirements	Functional (which data, which consumers, which queries/SLAs) + non-functional (volume, velocity, variety; freshness, availability, durability, consistency; budget; compliance).
2 · Estimate	Back-of-envelope: throughput, bytes/day & retention → storage, read QPS, rough cost. Sanity-checks every later choice.
3 · Architecture	Sketch sources → ingestion → storage → processing → serving, with orchestration and catalog/governance as cross-cutting layers.
4 · Decisions / trade-offs	The forks the interview names — batch vs streaming, lake vs warehouse vs lakehouse, CDC strategy, build vs buy — each justified against the requirements.
5 · Deep dives	Follow the interviewer's pulls: CDC correctness, hotspots/skew, exactly-once, schema evolution, failure.
6 · Evolution	Scale 10×, failure modes & recovery, migration, cost control.

Companion guide

This is the architect-and-choose guide. Its sibling, Production Data Pipelines, is the build-and-operate-robustly guide — late-arriving data, incremental processing, reconciliation, and orchestration. They share a topic and cross-reference each other; if your loop includes both a "systems design" round and a "production pipeline" round, do both.

Capstone: The Design Session Drive the 60-min whiteboard, worked end-to-end Drill: Practice Questions ~30 Q&A · hide-show answers Execution: Day-Of Tactics Method, whiteboard moves, red flags

Study pathsPick the plan that fits the time you have

a week

Day 1: 00, 01 (orient + the method) → 02 (requirements & sizing).
Day 2: 03 (ingestion) — a named topic.
Day 3: 04 (CDC) — a named topic, go deep.
Day 4: 05 (batch vs streaming) + 06 (lakes vs warehouses).
Day 5: 07 (build vs buy) + 08 (scalability & reliability).
Day 6: 09 (modeling & governance) + 10 (do the worked design session on a whiteboard yourself).
Day 7: Drill 11. Reread 12. Sleep.

two days

00, 01, 02, then the named topics and trade-offs — 03 (ingestion), 04 (CDC), 05 (batch/stream), 06 (lake/warehouse), 07 (build/buy) — then 10 (design session), drill 11, and 12. Skim 08, 09.

an evening

01 (the method + how it's graded), 04 (CDC), the decision matrices in 05, 06, and 07, and the framework in 10. Those are the load-bearing pieces of the whole conversation.

Section A · Orient

00 Start Here What this interview is, how it differs from a pipeline-build round, the design method that runs through the guide, reading order, and a study schedule. Read first. 01 The Design Interview & the Method How a 60-minute data systems design discussion is graded, the six-step method in depth, how to defend a trade-off, and how to whiteboard a platform so the interviewer follows you.

Section B · Building Blocks

02 Requirements & Estimation Pin the functional and non-functional requirements before drawing — freshness vs latency vs durability vs consistency — and size the system with back-of-envelope math that constrains every later choice. 03 Data Ingestion Pipelines Ingestion architecture at scale — sources, push vs pull, the log/queue backbone (Kafka/Kinesis/Pub-Sub), connectors, delivery guarantees, schema registry, backpressure, and dead-letter. A named topic. Named topic 04 Change Data Capture (CDC) CDC strategies compared — query-based vs trigger vs log-based (Debezium/WAL), snapshot-plus-stream, handling deletes/ordering/schema change, landing to a lake/warehouse, and the outbox alternative. A named topic. Named topic

Section C · The Big Trade-offs

05 Batch vs Streaming The central fork — batch, micro-batch, streaming; lambda vs kappa; engines; and a decision matrix keyed to freshness SLA vs cost, complexity, and correctness. A named trade-off. Decision matrix 06 Lakes, Warehouses & Lakehouses Warehouse vs lake vs lakehouse; open table formats (Iceberg/Delta/Hudi); columnar layout, partitioning, medallion, serving layers; and a decision matrix on cost, openness, performance, and workload. A named trade-off. Decision matrix 07 Build vs Buy The TCO framework, the managed-SaaS-to-in-house spectrum, the modern data stack, component-by-component build-vs-buy, lock-in, and a scorecard with a crisp heuristic. A named trade-off. Decision matrix

Section D · Cross-Cutting Concerns

08 Scalability & Reliability Partitioning and the hot-partition problem, replication and durability, fault tolerance and backpressure, exactly-once, the CAP/PACELC framing for data platforms, and multi-region / DR. 09 Modeling, Schema Evolution & Governance Dimensional vs Data Vault vs wide tables, schema evolution and compatibility, data contracts, catalog and lineage, PII/security/compliance, and cost governance — what makes a platform usable and trustworthy.

Section E · Putting It Together

10 The Design Session The capstone. The six-step method walked end-to-end on a full platform scenario — sizing, an architecture diagram, and every trade-off (CDC, batch/stream, lake/warehouse, build/buy) defended against requirements. 11 Practice Questions ~30 Q&A across requirements, ingestion, CDC, the trade-offs, scalability, and governance — each with a model answer and the decision heuristic. Drill mode hides the answers. Drill mode 12 Day-Of Tactics The method, whiteboarding moves for platform diagrams, how to defend trade-offs, questions to ask, red flags, and recovery patterns. Reread the morning of.