What's the query volume - peak QPS, geographic distribution?

1B documents in the corpus, 100K queries/sec at peak, 10M document updates/day. Multi-region read replicas. ~5 TB raw corpus, ~25 TB indexed (with replicas).

Strict latency budget - typical product search is p99 < 100ms; is this fixed?

Search p99 < 100ms (the search bar is on the user-facing path; every additional 100ms drops conversion measurably). Autocomplete p99 < 50ms (must keep up with typing). Index lag p95 < 60 seconds.

gitGood.dev

Design Search and Autocomplete (Elasticsearch-style)

Hard Premium

Inverted indexes, BM25 ranking, prefix tries, and the p99 < 100ms latency budget that drives every architectural choice.

The problem

Design a search system that serves two related queries: (1) full-text search over a document corpus (products, articles, jobs, code), returning ranked results in under 100ms p99; (2) autocomplete suggestions as the user types, returning the top 10 candidates in under 50ms p99. Both must handle continuous index updates as the underlying corpus changes.

This problem is graded on three things: the choice of inverted index structure and how it shards, the ranking pipeline (lexical → learned), and the latency engineering that makes p99 < 100ms achievable. Strong candidates open with the latency budget and design backwards from it.

Clarifying questions

Asking these before diving into a solution is the difference between a "hire" and a "no signal" rating. Pick the questions whose answers would change your design.

→What's the corpus - documents (text-heavy), products (structured), code (tokens), or mixed?
→What's the corpus size - millions, billions, or trillions of documents?
→Index update frequency - real-time (search-as-you-publish), near-real-time (minutes), or batch (hours)?
→What's the query volume - peak QPS, geographic distribution?
→Ranking requirements - pure lexical (BM25), learned-to-rank, personalized, or combination?
→Are there filters and facets (price range, category, date)? How dynamic?
→Spell correction, synonyms, multi-language support?
→Strict latency budget - typical product search is p99 < 100ms; is this fixed?

Requirements

Functional requirements

·GET /search?q=<query>&filters=...&page=N - returns ranked results with pagination
·GET /autocomplete?prefix=<chars> - returns top 10 query completions
·Index updates: insert, update, delete documents
·Filtering and faceted navigation (price ranges, categories, brand)
·Spell correction ("did you mean...")
·Personalization signals (user history, location)
·Multi-language tokenization and stemming

Non-functional requirements

Scale: 1B documents in the corpus, 100K queries/sec at peak, 10M document updates/day. Multi-region read replicas. ~5 TB raw corpus, ~25 TB indexed (with replicas).
Latency: Search p99 < 100ms (the search bar is on the user-facing path; every additional 100ms drops conversion measurably). Autocomplete p99 < 50ms (must keep up with typing). Index lag p95 < 60 seconds.
Availability: 99.95% for queries. Brief query outages degrade UX but don't lose data. Index updates can tolerate minutes of lag; complete outages of indexing are recoverable from the source of truth.
Consistency: Eventually consistent. A document just published can take seconds to appear in search results. Users tolerate this; the alternative (synchronous indexing on the write path) ruins write throughput.

The full walkthrough covers

Architecture components

·Document ingester
·Analyzer / tokenizer
·Index builder
·Shard / segment manager
·Query router (coordinator)
·Shard (query node)
·Re-ranking service
·Autocomplete service
·Spell-correction service

Deep dives

Inverted index: the data structure that makes search fast
BM25, TF-IDF, and learned-to-rank
Sharding strategies and the fan-out cost
Autocomplete: the trie, the FST, and the popularity layer
Indexing pipeline: real-time vs near-real-time vs batch
Filters, facets, and the relevance vs precision dance

What's in the full walkthrough

Capacity estimation · architecture with all 9 components explained · 6 deep dives · trade-off analysis · 8 common follow-up questions

Subscribe to Continue

Get full access to gitGood.dev with a subscription.

Start Your 10-Day Trial

Try everything free for 10 days. Cancel anytime.

1,950+ practice questions
AI mock interviews (3/day)
Coding challenges with real execution
Verified skill badges
146 certification guides, 26 with practice exams

View Plans

10-day trial, then $5/mo - or $40/yr ($3.33/mo, save 33%). Cancel anytime.

Companies that test this topic

Practice in interview format

Reading is the floor. The interview signal is in walking through this live with someone probing follow-ups. Use the AI mock interview to practice talking through requirements, architecture, and trade-offs out loud.

Start an AI mock interview →

Design Search and Autocomplete (Elasticsearch-style)

Hard Premium

Inverted indexes, BM25 ranking, prefix tries, and the p99 < 100ms latency budget that drives every architectural choice.

The problem

Clarifying questions

Asking these before diving into a solution is the difference between a "hire" and a "no signal" rating. Pick the questions whose answers would change your design.

→What's the corpus - documents (text-heavy), products (structured), code (tokens), or mixed?
→What's the corpus size - millions, billions, or trillions of documents?
→Index update frequency - real-time (search-as-you-publish), near-real-time (minutes), or batch (hours)?
→What's the query volume - peak QPS, geographic distribution?
→Ranking requirements - pure lexical (BM25), learned-to-rank, personalized, or combination?
→Are there filters and facets (price range, category, date)? How dynamic?
→Spell correction, synonyms, multi-language support?
→Strict latency budget - typical product search is p99 < 100ms; is this fixed?

Requirements

Functional requirements

·GET /search?q=<query>&filters=...&page=N - returns ranked results with pagination
·GET /autocomplete?prefix=<chars> - returns top 10 query completions
·Index updates: insert, update, delete documents
·Filtering and faceted navigation (price ranges, categories, brand)
·Spell correction ("did you mean...")
·Personalization signals (user history, location)
·Multi-language tokenization and stemming

Non-functional requirements

Scale: 1B documents in the corpus, 100K queries/sec at peak, 10M document updates/day. Multi-region read replicas. ~5 TB raw corpus, ~25 TB indexed (with replicas).
Latency: Search p99 < 100ms (the search bar is on the user-facing path; every additional 100ms drops conversion measurably). Autocomplete p99 < 50ms (must keep up with typing). Index lag p95 < 60 seconds.
Availability: 99.95% for queries. Brief query outages degrade UX but don't lose data. Index updates can tolerate minutes of lag; complete outages of indexing are recoverable from the source of truth.
Consistency: Eventually consistent. A document just published can take seconds to appear in search results. Users tolerate this; the alternative (synchronous indexing on the write path) ruins write throughput.

The full walkthrough covers

Architecture components

·Document ingester
·Analyzer / tokenizer
·Index builder
·Shard / segment manager
·Query router (coordinator)
·Shard (query node)
·Re-ranking service
·Autocomplete service
·Spell-correction service

Deep dives

Inverted index: the data structure that makes search fast
BM25, TF-IDF, and learned-to-rank
Sharding strategies and the fan-out cost
Autocomplete: the trie, the FST, and the popularity layer
Indexing pipeline: real-time vs near-real-time vs batch
Filters, facets, and the relevance vs precision dance

What's in the full walkthrough

Capacity estimation · architecture with all 9 components explained · 6 deep dives · trade-off analysis · 8 common follow-up questions

Subscribe to Continue

Get full access to gitGood.dev with a subscription.

Start Your 10-Day Trial

Try everything free for 10 days. Cancel anytime.

1,950+ practice questions
AI mock interviews (3/day)
Coding challenges with real execution
Verified skill badges
146 certification guides, 26 with practice exams

View Plans

10-day trial, then $5/mo - or $40/yr ($3.33/mo, save 33%). Cancel anytime.

Companies that test this topic

Practice in interview format

Start an AI mock interview →

Design Search and Autocomplete (Elasticsearch-style)

The problem

Clarifying questions

Requirements

Functional requirements

Non-functional requirements

The full walkthrough covers

Architecture components

Deep dives

What's in the full walkthrough

Subscribe to Continue

Start Your 10-Day Trial

Related system design topics

Distributed Cache

Web Crawler

Companies that test this topic

Practice in interview format

Design Search and Autocomplete (Elasticsearch-style)

The problem

Clarifying questions

Requirements

Functional requirements

Non-functional requirements

The full walkthrough covers

Architecture components

Deep dives

What's in the full walkthrough

Subscribe to Continue

Start Your 10-Day Trial

Related system design topics

Distributed Cache

Web Crawler

Companies that test this topic

Practice in interview format