Listening: System Design Interview

In system design interviews, asking clarifying questions before designing is the expected and correct behaviour. It signals that the candidate thinks before they code — a key engineering virtue.

Clarifying question vocabulary:

• "before jumping into the design" — deliberate pause before starting; shows discipline
• "read-heavy vs write-heavy" — a fundamental question: are more operations reads (URL redirects) or writes (URL creations)? This shapes the entire architecture.
• "what scale are we talking" — prompting the interviewer to define the traffic and data volume requirements
• "do we need analytics" — asking about features that may not be in scope; confirms what to design

What interviewers look for: A candidate who jumps straight to a solution without clarifying is a red flag. It suggests they make assumptions in real projects too. Structured clarification first = senior engineering behaviour.

"Back-of-the-envelope calculation" (sometimes abbreviated BOTE) is an idiom for a quick, rough estimate using simple arithmetic and round numbers. The phrase comes from the idea of doing a calculation on the back of an envelope — informally, without formal tools.

Estimation vocabulary in interviews:

• "rough" — approximate, not precise; openly acknowledged uncertainty
• "let's call it X" — rounding to a convenient number; signals intentional simplification
• "add some headroom" — build in a safety margin; engineer for more than the minimum
• "read-to-write ratio" — the proportion of read operations vs write operations; drives caching and database design decisions

Why this matters: Interviewers assess whether candidates can reason quantitatively about scale. Arriving at "100:1 read-heavy" tells the interviewer the candidate will correctly prioritise read optimisation (caching, CDN) in the design.

"Graceful degradation" is a core resilience principle: when a component fails, the system should continue offering as much functionality as possible rather than going completely offline.

Failure discussion vocabulary:

• "degrade gracefully" — maintain partial functionality during failure; opposite of "fail catastrophically"
• "read replica" — a copy of the database used for read queries; can serve reads even if the primary is unavailable
• "failover" — automatic switching to a backup system when the primary fails
• "window of unavailability" — a brief period when the service cannot accept requests; acknowledged and accepted as a trade-off
• "message queue" — a buffer that stores write requests until the primary database recovers; enables eventual consistency

"Let me think through this" is excellent interview language — it signals that the candidate thinks out loud (valued by interviewers) rather than rushing to a potentially wrong answer.