Interview Concepts

A glossary of the concepts interviews are built on, in plain English — what each one is, why interviewers probe it, and where to practice it. Covers the distributed-systems vocabulary asked at FAANG (Facebook/Meta, Apple, Amazon, Netflix, Google) and the LLM infrastructure terms asked at AI-first companies like Anthropic, OpenAI, and Nvidia.

Concepts: 47
Areas: 3

System Design

BackpressureBackpressure is a system's way of telling upstream senders to slow down when it cannot keep up, instead of silently queueing work until it collapses.Definition + interview angle
CAP TheoremThe CAP theorem states that during a network partition, a distributed system must choose between consistency and availability — it cannot keep both.Definition + interview angle
Circuit BreakerA circuit breaker stops calling a failing downstream service and fails fast for a cooldown period, preventing retry storms and cascading failures.Definition + interview angle
Consistent HashingConsistent hashing places nodes and keys on a ring so that adding or removing a node remaps only ~1/N of the keys, instead of nearly all of them.Definition + interview angle
Content Delivery Network (CDN)A CDN is a network of geographically distributed edge servers that cache content close to users, cutting latency and shielding the origin from load.Definition + interview angle
Eventual ConsistencyEventual consistency guarantees that replicas converge to the same value once writes stop — but says nothing about how stale a read may be in the meantime.Definition + interview angle
Fan-Out (Write vs Read)Fan-out is how a feed distributes one item to many followers — on write (push to every follower's feed) or on read (assemble at request time).Definition + interview angle
IdempotencyAn idempotent operation produces the same result whether applied once or many times — the property that makes retries safe in distributed systems.Definition + interview angle
Leader ElectionLeader election is how a cluster agrees on one node to coordinate work, and how it safely replaces that node when it fails or becomes unreachable.Definition + interview angle
Load BalancerA load balancer distributes incoming traffic across multiple servers; L4 balancers route on IPs and ports, L7 balancers route on HTTP paths and headers.Definition + interview angle
Message QueueA message queue decouples a fast producer from a slow consumer: work is enqueued in milliseconds and processed asynchronously, so bursts queue up instead of timing out.Definition + interview angle
NALSD (Non-Abstract Large System Design)NALSD is Google's SRE approach to system design: iterate a design into concrete numbers — machines, disks, QPS — proving it works at scale, not just in boxes.Definition + interview angle
Quorum (R + W > N)A quorum system requires R read and W write acknowledgments from N replicas with R + W > N, so every read overlaps the latest write in at least one replica.Definition + interview angle
Rate LimitingRate limiting caps how many requests a client may make per time window; the token bucket algorithm allows short bursts while enforcing an average rate.Definition + interview angle
Service Level Objective (SLO)An SLO is a target for a reliability metric over a window — for example 99.9% of requests under 300ms in 28 days — and the error budget it implies.Definition + interview angle
ShardingSharding splits one large dataset across multiple database nodes by a shard key, so each node stores and serves only a fraction of the data.Definition + interview angle
Tail Latency (p99)Tail latency is the latency of a system's slowest percentiles (p95, p99). With many backend calls per page, a large share of users hit at least one p99-slow call.Definition + interview angle

LLM & AI Engineering

Databases & Caching

Interview vocabulary FAQ

What terms do system design interviews expect you to know?

The recurring vocabulary falls into three groups: distributed-systems fundamentals (CAP theorem, consistent hashing, sharding, quorums, idempotency), storage and caching (indexes, replication lag, cache-aside, write-ahead logs), and — increasingly since 2024 — LLM infrastructure terms (KV cache, RAG, embeddings, quantization). Interviewers rarely ask for definitions directly; they expect you to use the terms correctly while designing and to reason with the concepts behind them.

What is NALSD in system design interviews?

NALSD stands for Non-Abstract Large System Design, the methodology used by Google's Site Reliability Engineering organization. It requires iterating a design past abstract boxes into concrete numbers — machine counts, disk throughput, queries per second per replica — until the arithmetic proves the design works at scale. NALSD-style rounds are standard in Google SRE interview loops, and the numbers-first style increasingly appears in senior system design interviews at other companies.

Is memorizing definitions enough to pass a system design interview?

No. Interviews test whether you can apply the concepts under follow-up pressure: choosing a shard key and defending what it makes expensive, running the quorum overlap argument for your replication settings, or putting numbers on a cache. Use these entries to close vocabulary gaps, then practice producing designs — each entry here links to interactive questions and quizzes where the concept is exercised rather than recited.

Vocabulary is the floor. Practice is the interview.

Every term here links into 140+ interactive system design and coding questions where you apply the concept — clarify requirements, make the tradeoffs, run the numbers.

Browse all questions Test yourself with a quiz