Search functionality design in LLD - System Design Exercise

Design: Search Functionality System

Design covers search query processing, indexing, ranking, and autocomplete. Does not cover user authentication or data ingestion pipelines.

Functional Requirements

FR1: Allow users to search for items by keywords

FR2: Support typo tolerance and partial matches

FR3: Return results ranked by relevance

FR4: Handle 1000 concurrent search requests

FR5: Provide search results within 300ms latency

FR6: Support filtering results by categories

FR7: Allow autocomplete suggestions as user types

Non-Functional Requirements

NFR1: System must be available 99.9% of the time

NFR2: Search index must update within 5 minutes of data changes

NFR3: Support up to 10 million searchable items

NFR4: Use scalable and cost-effective technologies

Think Before You Design

Questions to Ask

❓ Question 1

❓ Question 2

❓ Question 3

❓ Question 4

❓ Question 5

❓ Question 6

Key Components

Search index storage (e.g., inverted index)

Query parser and processor

Ranking and scoring module

Autocomplete suggestion engine

Cache layer for popular queries

Data update and index refresh mechanism

Design Patterns

Inverted index for fast keyword lookup

Trie or prefix tree for autocomplete

TF-IDF or BM25 ranking algorithms

Caching frequently searched queries

Batch and incremental index updates

Reference Architecture

User
  |
  v
Search API Server <--> Cache Layer <--> Search Index Storage
  |
  v
Autocomplete Service
  |
  v
Data Update Service --> Index Updater --> Search Index Storage

Components

Search API Server

Node.js or Python REST API

Receives search requests, parses queries, and returns ranked results

Cache Layer

Redis

Stores results of popular queries to reduce latency

Search Index Storage

Elasticsearch or Apache Lucene

Stores inverted index and supports fast keyword search and ranking

Autocomplete Service

Trie data structure in memory or Elasticsearch completion suggester

Provides real-time autocomplete suggestions as user types

Data Update Service

Batch jobs or streaming pipeline

Processes new or updated items and triggers index refresh

Index Updater

Elasticsearch bulk API or custom indexer

Updates the search index with new data within 5 minutes

Request Flow

1. User sends search query to Search API Server

2. Search API Server checks Cache Layer for cached results

3. If cache miss, Search API Server queries Search Index Storage

4. Search Index Storage returns ranked results based on query

5. Search API Server returns results to user and caches them

6. For autocomplete, user input is sent to Autocomplete Service

7. Autocomplete Service returns suggestions in real-time

8. Data Update Service processes new data and sends to Index Updater

9. Index Updater refreshes Search Index Storage within 5 minutes

Database Schema

Entities: - Item: id (PK), title, description, category, tags, updated_at - SearchIndex: inverted index structure mapping keywords to item ids - AutocompleteTrie: prefix tree nodes storing partial keywords Relationships: - Items are indexed into SearchIndex for keyword lookup - AutocompleteTrie built from item titles and keywords for suggestions

Scaling Discussion

Bottlenecks

Search API Server CPU and memory limits under high concurrency

Cache Layer memory capacity for popular queries

Search Index Storage disk I/O and query throughput

Index update latency with large data volumes

Autocomplete service response time with large prefix sets

Solutions

Scale Search API Server horizontally behind load balancer

Use distributed cache clusters and eviction policies

Shard search index across multiple nodes and use replicas

Implement incremental and parallel index updates

Optimize autocomplete data structures and cache hot prefixes

Interview Tips

Time: Spend 10 minutes clarifying requirements and constraints, 20 minutes designing architecture and data flow, 10 minutes discussing scaling and trade-offs, 5 minutes summarizing.

Clarify search use cases and data update frequency

Explain choice of inverted index and autocomplete structures

Describe caching strategy to reduce latency

Discuss ranking algorithms and relevance tuning

Address scaling challenges and solutions

Highlight trade-offs between freshness and performance