Design a Concurrent Hash Map

Striped locking or segments for parallel access.

Advanced14 min read

Read these first

Design In-Process Pub-Sub

Why this matters

put/get/remove with expected O(1) average time
Scope — what "Segments" covers inside Concurrent Hash Map and what it does not
Scope — what "Resize" covers inside Concurrent Hash Map and what it does not
Scope — what "vs ConcurrentHashMap" covers inside Concurrent Hash Map and what it does not
In-memory systems test data structures + thread safety together.

Segment 0

Segment 1

Segment 2

Lock striping: each segment has its own lock for parallel writes.

CampusOS context

CampusOS is the fictional smart-campus platform used across this track. Treat this module as a real product slice:

Context

Belongs to CampusOS — same naming and design vocabulary as parking, library, and booking modules
Scope is one JVM — persistence is in-memory or behind a repository interface
Goal is a class diagram + core Java that an interviewer can follow in 35–45 minutes
Extension paragraph at the end links to System Design when scale exceeds one process

Requirements snapshot

Functional requirements

put/get/remove with expected O(1) average time
Thread-safe under concurrent writers
resize when load factor exceeded
Optional: computeIfAbsent atomic operation

Non-functional requirements

Single JVM, in-process design for this LLD article
Core flows testable with in-memory repositories
Extension points for new rules without editing central god classes
Clear public API on facade/service layer
CampusOS extension links to System Design when scale demands distribution

Overview

This CampusOS module designs Concurrent Hash Map as a production-ready in-process service. Read requirements point-by-point before drawing classes.

Requirements

Points to cover

put/get/remove with expected O(1) average time
Thread-safe under concurrent writers
resize when load factor exceeded
Optional: computeIfAbsent atomic operation

Segments

Points to cover

Scope — what "Segments" covers inside Concurrent Hash Map and what it does not
Classes — name 2–4 classes/interfaces responsible for this slice
Flow — step-by-step: trigger → validation → state change → side effect
Edge cases — invalid input, concurrent access, empty state, timeout
Pattern — Strategy, State, Observer, or Facade if it fits naturally
Test hook — one scenario an interviewer will ask you to walk through

Resize

Points to cover

Scope — what "Resize" covers inside Concurrent Hash Map and what it does not
Classes — name 2–4 classes/interfaces responsible for this slice
Flow — step-by-step: trigger → validation → state change → side effect
Edge cases — invalid input, concurrent access, empty state, timeout
Pattern — Strategy, State, Observer, or Facade if it fits naturally
Test hook — one scenario an interviewer will ask you to walk through

vs ConcurrentHashMap

Points to cover

Scope — what "vs ConcurrentHashMap" covers inside Concurrent Hash Map and what it does not
Classes — name 2–4 classes/interfaces responsible for this slice
Flow — step-by-step: trigger → validation → state change → side effect
Edge cases — invalid input, concurrent access, empty state, timeout
Pattern — Strategy, State, Observer, or Facade if it fits naturally
Test hook — one scenario an interviewer will ask you to walk through

Java

// CampusOS — ConcurrentHashMap facade (keep thin)
public final class ConcurrentHashMapService {
    private final ConcurrentHashMapRepository repository;

    public ConcurrentHashMapService(ConcurrentHashMapRepository repository) {
        this.repository = repository;
    }

    /** Orchestrate: validate → domain rules → persist → return result */
    public Result handle(Request request) {
        request.validate();
        var domain = DomainModel.from(request);
        domain.applyRules();
        return repository.save(domain);
    }
}

Quick recall

Everything you need if you only revisit this box.

Requirements — review this slice before mock interviews.
Segments — review this slice before mock interviews.
Resize — review this slice before mock interviews.
vs ConcurrentHashMap — review this slice before mock interviews.

CampusOS extension

When this outgrows one JVM

Shard inventory or state by campus ID or geographic region
Push notifications, payments, and search to dedicated services (System Design track)
Use message queues for async side effects (email, analytics)
The class diagram you drew here becomes one box on the HLD architecture diagram

Test yourself

Answer these before moving on — recall is what makes it stick.