Chat & Messaging — System Design Academy

Complete System Design Tutorial

The design separates the latency-sensitive online path, the correctness-critical durable transition, and asynchronous projection or repair work. Each chapter below explains one decision shown in the diagrams.

1. Requirements and scope

Functional requirements

Execute the primary chat & messaging workflow and return an observable result.
Expose the concrete operations shown below with stable retry identities.
Allow operators to inspect, disable, repair, or replay ambiguous state.
Keep derived analytics and notifications off the correctness-critical path.

Non-functional requirements

Design for 50M concurrent sockets · 5B messages/day · p99 < 300 ms.
Protect the named business invariant under concurrency and partial failure.
Scale by an explicit aggregate/partition key without one global mutex.
Degrade intentionally when cache, queue, store, or regional ownership is unavailable.

Primary invariant: accepted events remain ordered within their conversation/session and survive reconnects.

↑ top

2. Capacity assumptions

State assumptions before choosing technology. The supplied workload is 50M concurrent sockets · 5B messages/day · p99 < 300 ms. Convert it into peak throughput, retained records, bandwidth, and failure budget; then connect each number to a component.

Dimension	Working estimate	Design implication
Traffic	50M concurrent sockets · 5B messages/day · p99 < 300 ms	Separate online compute, durable state, and asynchronous consumers.
Peak factor	5–10× average	Admission control, autoscaling headroom, and hot-key protection.
Stored state	write rate × retention × record bytes	Partition count, indexes, compaction, backup, and archival tier.
Network	peak operations × payload size	Batching, compression, edge placement, and egress budget.
Availability	99.9–99.99%	Multi-zone replicas, tested failover, and dependency error budgets.
Recovery	RPO/RTO by invariant	Ack boundary, replay retention, reconciliation, and operator runbooks.

Do not cargo-cult the numbers: show the arithmetic, name the uncertain variable, and say which production metric will replace the interview estimate.

↑ top

3. API and data model

WS SEND_MESSAGE(client_msg_id, conversation_id, body)

GET /v1/sync?cursor=resume missed events

POST /v1/conversationscreate membership

PUT /v1/receipts/{conversation_id}advance read cursor

Authoritative records and indexes

message(conversation_id, sequence, message_id, sender_id, body_ref)

device_cursor(user_id, device_id, partition, sequence)

membership(conversation_id, user_id, role, version)

receipt(conversation_id, user_id, delivered_seq, read_seq)

Command and event envelopes

# Command envelope
operation: WS SEND_MESSAGE(client_msg_id, conversation_id, body)
request_id: req_01J...
idempotency_key: idem_7f2...
aggregate_key: chat-system_123
expected_version: 17
payload: {...}

# Durable event
event_id: evt_01J...
aggregate_key: chat-system_123
event_type: chat_system_committed
schema_version: 3
trace_id: tr_8ac...

The authoritative aggregate and idempotency record share a transactional or conditional-write boundary. Event streams, search indexes, caches, analytics, and dashboards are rebuildable projections.

↑ top

4. Why this architecture and what it costs

The boxes are not a shopping list. Each boundary exists because it owns a different invariant, scaling dimension, data lifecycle, security policy, or failure domain. The table makes those choices explicit and shows when the design should remain simpler.

Component	Owned responsibility	Why separate it	Rejected alternative	Cost and evolution trigger
Mobile / Web	Owns one bounded domain decision, transition, or operational capability. In this design it participates in: Authenticate the socket and resolve its gateway in the session directory.	Separates a distinct scale, ownership, policy, or failure boundary. It prevents the external caller from also owning the contract and failure behavior of Anycast Edge.	Keep this behavior inside an adjacent service. Prefer that simpler option until independent scale, ownership, policy, or failure isolation is measurable. Here that simpler choice would couple “Gateway dies after log append before acknowledgment” to the external caller.	Costs an interface, deployment, telemetry, ownership, and failure mode. Keep it in-process first; extract when independent scale or reliability pays for the boundary. Measure the latency, saturation, lag, or mismatch at the Mobile / Web boundary before extracting or scaling it.
Anycast Edge	Owns authentication, validation, quotas, deadlines, and routing. In this design it participates in: Route a send to the conversation partition and recheck membership version.	Separates a distinct scale, ownership, policy, or failure boundary. It prevents Mobile / Web from also owning the contract and failure behavior of WebSocket Gateway.	Put protocol, authentication, quotas, and business transitions in one process. Acceptable at low scale, but it couples external traffic risk to domain releases. Here that simpler choice would couple “Large group fan-out overwhelms online delivery” to Mobile / Web.	Costs another hop, policy distribution, and a high-blast-radius fleet. Introduce it when many clients or services need one protocol and protection boundary. Measure the latency, saturation, lag, or mismatch at the Anycast Edge boundary before extracting or scaling it.
WebSocket Gateway	Owns authentication, validation, quotas, deadlines, and routing. In this design it participates in: Deduplicate client_msg_id, assign the next partition sequence, and durably append.	Separates a distinct scale, ownership, policy, or failure boundary. It prevents Anycast Edge from also owning the contract and failure behavior of Session Directory.	Put protocol, authentication, quotas, and business transitions in one process. Acceptable at low scale, but it couples external traffic risk to domain releases. Here that simpler choice would couple “Membership revoked during an in-flight send” to Anycast Edge.	Costs another hop, policy distribution, and a high-blast-radius fleet. Introduce it when many clients or services need one protocol and protection boundary. Measure the latency, saturation, lag, or mismatch at the WebSocket Gateway boundary before extracting or scaling it.
Session Directory	Owns one bounded domain decision, transition, or operational capability. In this design it participates in: Acknowledge the sender, then fan out to online gateways and offline inboxes.	Separates a distinct scale, ownership, policy, or failure boundary. It prevents WebSocket Gateway from also owning the contract and failure behavior of Auth / Membership.	Keep this behavior inside an adjacent service. Prefer that simpler option until independent scale, ownership, policy, or failure isolation is measurable. Here that simpler choice would couple “Two devices race receipt cursors” to WebSocket Gateway.	Costs an interface, deployment, telemetry, ownership, and failure mode. Keep it in-process first; extract when independent scale or reliability pays for the boundary. Measure the latency, saturation, lag, or mismatch at the Session Directory boundary before extracting or scaling it.
Auth / Membership	Owns one bounded domain decision, transition, or operational capability. In this design it participates in: Devices resume from cursors; receipt updates coalesce to monotonic sequence advances.	Separates a distinct scale, ownership, policy, or failure boundary. It prevents Session Directory from also owning the contract and failure behavior of Conversation Router.	Embed policy in every caller. Rules then drift by deployment version and emergency changes require fleet-wide redeployment. Here that simpler choice would couple “Regional partition owner becomes unavailable” to Session Directory.	Costs an interface, deployment, telemetry, ownership, and failure mode. Keep it in-process first; extract when independent scale or reliability pays for the boundary. Measure the latency, saturation, lag, or mismatch at the Auth / Membership boundary before extracting or scaling it.
Conversation Router	Owns one bounded domain decision, transition, or operational capability. In this design it participates in: Authenticate the socket and resolve its gateway in the session directory.	Separates a distinct scale, ownership, policy, or failure boundary. It prevents Auth / Membership from also owning the contract and failure behavior of Message Sequencer.	Keep this behavior inside an adjacent service. Prefer that simpler option until independent scale, ownership, policy, or failure isolation is measurable. Here that simpler choice would couple “Gateway dies after log append before acknowledgment” to Auth / Membership.	Costs an interface, deployment, telemetry, ownership, and failure mode. Keep it in-process first; extract when independent scale or reliability pays for the boundary. Measure the latency, saturation, lag, or mismatch at the Conversation Router boundary before extracting or scaling it.
Message Sequencer	Owns one bounded domain decision, transition, or operational capability. In this design it participates in: Route a send to the conversation partition and recheck membership version.	Separates a distinct scale, ownership, policy, or failure boundary. It prevents Conversation Router from also owning the contract and failure behavior of Partitioned Message Log.	Keep this behavior inside an adjacent service. Prefer that simpler option until independent scale, ownership, policy, or failure isolation is measurable. Here that simpler choice would couple “Large group fan-out overwhelms online delivery” to Conversation Router.	Costs an interface, deployment, telemetry, ownership, and failure mode. Keep it in-process first; extract when independent scale or reliability pays for the boundary. Measure the latency, saturation, lag, or mismatch at the Message Sequencer boundary before extracting or scaling it.
Partitioned Message Log	Preserves durable ordering, replay, and consumer checkpoints. In this design it participates in: Deduplicate client_msg_id, assign the next partition sequence, and durably append.	Separates a distinct scale, ownership, policy, or failure boundary. It prevents Message Sequencer from also owning the contract and failure behavior of Inbox Fanout.	Call every downstream consumer synchronously. That couples user latency and availability to analytics, notification, indexing, and audit dependencies. Here that simpler choice would couple “Membership revoked during an in-flight send” to Message Sequencer.	Costs eventual consistency, schema evolution, replay tooling, idempotent consumers, and lag operations. Earn it when bursts or downstream failure must not block acknowledgment. Measure the latency, saturation, lag, or mismatch at the Partitioned Message Log boundary before extracting or scaling it.
Inbox Fanout	Executes idempotent asynchronous work under lease and retry policy. In this design it participates in: Acknowledge the sender, then fan out to online gateways and offline inboxes.	Separates a distinct scale, ownership, policy, or failure boundary. It prevents Partitioned Message Log from also owning the contract and failure behavior of Push Service.	Perform the work before acknowledging the caller. This gives simpler consistency but unbounded tail latency and fragile retries. Here that simpler choice would couple “Two devices race receipt cursors” to Partitioned Message Log.	Costs leases, checkpoints, retries, DLQs, and lag operations. Separate once execution time or burstiness no longer fits the request deadline. Measure the latency, saturation, lag, or mismatch at the Inbox Fanout boundary before extracting or scaling it.
Push Service	Owns one bounded domain decision, transition, or operational capability. In this design it participates in: Devices resume from cursors; receipt updates coalesce to monotonic sequence advances.	Separates a distinct scale, ownership, policy, or failure boundary. It prevents Inbox Fanout from also owning the contract and failure behavior of Presence Store.	Keep this behavior inside an adjacent service. Prefer that simpler option until independent scale, ownership, policy, or failure isolation is measurable. Here that simpler choice would couple “Regional partition owner becomes unavailable” to Inbox Fanout.	Costs an interface, deployment, telemetry, ownership, and failure mode. Keep it in-process first; extract when independent scale or reliability pays for the boundary. Measure the latency, saturation, lag, or mismatch at the Push Service boundary before extracting or scaling it.
Presence Store	Owns authoritative versioned state or immutable evidence. In this design it participates in: Authenticate the socket and resolve its gateway in the session directory.	Defines the source of truth and commit boundary. It prevents Push Service from also owning the contract and failure behavior of Receipt Aggregator.	Let each service persist its own interpretation. That removes one clear commit boundary and makes retries, support, and reconciliation ambiguous. Here that simpler choice would couple “Gateway dies after log append before acknowledgment” to Push Service.	Costs schema migration, replication, backup, and shard ownership. Split storage only when access patterns, retention, or compliance boundaries materially differ. Measure the latency, saturation, lag, or mismatch at the Presence Store boundary before extracting or scaling it.
Receipt Aggregator	Owns one bounded domain decision, transition, or operational capability. In this design it participates in: Route a send to the conversation partition and recheck membership version.	Separates a distinct scale, ownership, policy, or failure boundary. It prevents Presence Store from also owning the contract and failure behavior of Media Store.	Keep this behavior inside an adjacent service. Prefer that simpler option until independent scale, ownership, policy, or failure isolation is measurable. Here that simpler choice would couple “Large group fan-out overwhelms online delivery” to Presence Store.	Costs an interface, deployment, telemetry, ownership, and failure mode. Keep it in-process first; extract when independent scale or reliability pays for the boundary. Measure the latency, saturation, lag, or mismatch at the Receipt Aggregator boundary before extracting or scaling it.
Media Store	Owns authoritative versioned state or immutable evidence. In this design it participates in: Deduplicate client_msg_id, assign the next partition sequence, and durably append.	Defines the source of truth and commit boundary. It prevents Receipt Aggregator from also owning the contract and failure behavior of Offline Sync API.	Let each service persist its own interpretation. That removes one clear commit boundary and makes retries, support, and reconciliation ambiguous. Here that simpler choice would couple “Membership revoked during an in-flight send” to Receipt Aggregator.	Costs schema migration, replication, backup, and shard ownership. Split storage only when access patterns, retention, or compliance boundaries materially differ. Measure the latency, saturation, lag, or mismatch at the Media Store boundary before extracting or scaling it.
Offline Sync API	Owns authentication, validation, quotas, deadlines, and routing. In this design it participates in: Acknowledge the sender, then fan out to online gateways and offline inboxes.	Separates a distinct scale, ownership, policy, or failure boundary. It prevents Media Store from also owning the contract and failure behavior of operators and repair tooling.	Put protocol, authentication, quotas, and business transitions in one process. Acceptable at low scale, but it couples external traffic risk to domain releases. Here that simpler choice would couple “Two devices race receipt cursors” to Media Store.	Costs another hop, policy distribution, and a high-blast-radius fleet. Introduce it when many clients or services need one protocol and protection boundary. Measure the latency, saturation, lag, or mismatch at the Offline Sync API boundary before extracting or scaling it.

Decision 01

Chosen boundary vs simpler monolith

Partition conversation logs by conversation_id to preserve per-chat ordering.

A single deployable with one transactional store is the right starting point for a small chat & messaging workload. The proposed boundary becomes justified when partition conversation logs by conversation_id to preserve per-chat ordering. Separating it contains the named failure “Gateway dies after log append before acknowledgment” without placing unrelated work on the critical path.

Decision 02

Correctness vs availability

Assign server sequence numbers, persist before acknowledgment, and fan out from a durable log.

The design coordinates only where accepted events remain ordered within their conversation/session and survive reconnects. Everywhere else it accepts bounded staleness or asynchronous propagation. The alternative—strong coordination for every projection—simplifies the mental model but increases tail latency and turns partitions into broad write unavailability.

Decision 03

Operational cost and evolution

Presence is ephemeral and eventually consistent; message history and membership are durable.

This choice adds dashboards, ownership, schema compatibility, capacity planning, and recovery runbooks. Start with the minimum topology that supports “Deduplicate client_msg_id, assign the next partition sequence, and durably append.” and extract another service only when measured throughput, blast radius, data lifecycle, or team ownership changes independently.

Why this design: the synchronous path contains only the decisions required to protect accepted events remain ordered within their conversation/session and survive reconnects. Durable projections, notifications, indexing, analytics, and repair are isolated so their load or failure cannot corrupt the acknowledged operation.

Correctness boundary

Partition conversation logs by conversation_id to preserve per-chat ordering.

Scale boundary

Assign server sequence numbers, persist before acknowledgment, and fan out from a durable log.

Operational boundary

Presence is ephemeral and eventually consistent; message history and membership are durable.

↑ top

5. Critical flow and algorithm

Authenticate the socket and resolve its gateway in the session directory.

Route a send to the conversation partition and recheck membership version.

Deduplicate client_msg_id, assign the next partition sequence, and durably append.

Acknowledge the sender, then fan out to online gateways and offline inboxes.

Devices resume from cursors; receipt updates coalesce to monotonic sequence advances.

Retry-safe mutation skeleton

identity = authenticate(request)
prior = idempotency.get(identity, request.key)
if prior: return prior.result

current = authoritative_store.read(partition_key)
next_state, events = domain.transition(current, request)
authoritative_store.compare_and_set(current.version, next_state)
idempotency.commit(request.key, next_state.result)
outbox.append(events)
return next_state.result

Exactly-once business effect: the network can duplicate requests. Stable identities, conditional state transitions, immutable event IDs, and reconciliation make the user-visible effect happen once.

↑ top

6. Caching, hot keys, and backpressure

Cache policy

Add a cache only for a measured repeated read; do not place mutable correctness state in a disposable layer.

Hot partition or tenant

Measure per-key and per-tenant skew. Replicate safe reads, isolate the offender, salt only when ordering permits, and enforce budgets.

Backpressure

Bound queues, buffers, connection pools, and concurrency. Shed low-priority work and scale from lag/oldest-item age rather than CPU alone.

Stampede and retry storms

Use single-flight/coalescing, jittered expiry, bounded retries, exponential backoff, and circuit breakers so one dependency failure does not multiply traffic.

Cache outage rule: never allow every miss to fall through simultaneously. Use local protection, request coalescing, circuit breakers, bounded pools, and controlled fallback.

↑ top

7. Partitioning and consistency

partition = stable_hash(primary_or_aggregate_key) → shard owner
write = conditional_update(expected_version)
read_projection = versioned and rebuildable

Strong boundary

accepted events remain ordered within their conversation/session and survive reconnects. Use a uniqueness constraint, serializable/conditional write, expected version, or fencing token at this boundary.

Eventual boundary

Analytics, notifications, search, read models, and operational dashboards may lag when freshness is observable and replay exists.

Operation	Consistency	Reason
Authoritative mutation	Conditional / serializable	Protect uniqueness, ownership, money, inventory, order, or legal state transition.
Read after own write	Primary, session token, or cache warm	The caller must observe the result it just committed.
Ordinary read projection	Bounded stale where safe	Replicas and caches improve latency when version/freshness is visible.
Analytics and operations	Eventual + replay	Freshness is less important than isolating the online path.

Route every mutation for one aggregate to one owner/partition epoch.
Use cursors rather than offsets for changing ordered collections.
Retain tombstones until replication and repair windows close.
Move shard ownership through versioned maps and dual-read verification.

↑ top

8. Failure drills and recovery

Failure	Runtime behavior	Recovery proof
Gateway dies after log append before acknowledgment	Stop unsafe progress, serve only explicitly safe degraded behavior, and recover from durable state.	Reconciliation returns zero invariant gaps; lag/replica health and user-visible SLO return to baseline.
Large group fan-out overwhelms online delivery	Stop unsafe progress, serve only explicitly safe degraded behavior, and recover from durable state.	Reconciliation returns zero invariant gaps; lag/replica health and user-visible SLO return to baseline.
Membership revoked during an in-flight send	Stop unsafe progress, serve only explicitly safe degraded behavior, and recover from durable state.	Reconciliation returns zero invariant gaps; lag/replica health and user-visible SLO return to baseline.
Two devices race receipt cursors	Resolve at source of truth with uniqueness, expected version, lease generation, or monotonic cursor.	Reconciliation returns zero invariant gaps; lag/replica health and user-visible SLO return to baseline.
Regional partition owner becomes unavailable	Stop unsafe progress, serve only explicitly safe degraded behavior, and recover from durable state.	Reconciliation returns zero invariant gaps; lag/replica health and user-visible SLO return to baseline.

Recovery checkpoint

# Recovery checkpoint
scenario: Gateway dies after log append before acknowledgment
operation_id: op_01J...
aggregate_key: chat-system_123
last_durable_version: 18
projection_checkpoint: partition-07@offset-482901
retry_policy: same_identity / bounded_backoff
reconciliation_status: pending
owner: Offline Sync API

Every dependency needs a timeout, retry budget, circuit policy, and explicit fail-open, fail-closed, stale, or read-only mode. “Highly available” is incomplete until degraded behavior is named.

↑ top

9. Security, regions, and observability

Security and abuse

Authenticate at ingress and authorize the exact resource/tenant on every operation.
Encrypt transport and durable data; tokenize or isolate sensitive material.
Rate-limit by tenant, actor, resource, and network; detect enumeration and anomalous fan-out.
Redact secrets and personal data before logs or analytical storage.
Audit privileged mutations, support access, policy changes, and forced recovery.

Signals that prove correctness

User-visible success rate and p50/p95/p99 latency.
Timeouts, circuit state, saturation, and replica health by dependency.
Queue lag, oldest-item age, retries, and dead-letter growth.
Conflict, duplicate-suppression, and reconciliation mismatch counts.
Business drill: Gateway dies after log append before acknowledgment — alert before the user-visible invariant is violated.

Regional evolution

Begin multi-zone with tested failover, backups, and reconciliation.
Add global stateless ingress and regional read projections.
Keep one write owner per aggregate or partition where coordination is expensive.
Move ownership with epochs/fencing; avoid silent simultaneous writers.
Use expand/contract schemas, canaries, shadow reads, and reversible cutovers.

↑ top

10. Interviewer follow-ups

How do group chats change fan-out?

Partition conversation logs by conversation_id to preserve per-chat ordering. Walk the concrete lifecycle step “Authenticate the socket and resolve its gateway in the session directory.” Then test it against “Gateway dies after log append before acknowledgment”. State the retry identity, durable checkpoint, recovery signal, and cost of the alternative.

How do you support multi-device sync?

Assign server sequence numbers, persist before acknowledgment, and fan out from a durable log. Walk the concrete lifecycle step “Route a send to the conversation partition and recheck membership version.” Then test it against “Large group fan-out overwhelms online delivery”. State the retry identity, durable checkpoint, recovery signal, and cost of the alternative.

What does exactly-once mean to a user?

Presence is ephemeral and eventually consistent; message history and membership are durable. Walk the concrete lifecycle step “Deduplicate client_msg_id, assign the next partition sequence, and durably append.” Then test it against “Membership revoked during an in-flight send”. State the retry identity, durable checkpoint, recovery signal, and cost of the alternative.

How do read receipts scale?

Partition conversation logs by conversation_id to preserve per-chat ordering. Walk the concrete lifecycle step “Acknowledge the sender, then fan out to online gateways and offline inboxes.” Then test it against “Two devices race receipt cursors”. State the retry identity, durable checkpoint, recovery signal, and cost of the alternative.

How do you handle a regional outage?

Assign server sequence numbers, persist before acknowledgment, and fan out from a durable log. Walk the concrete lifecycle step “Devices resume from cursors; receipt updates coalesce to monotonic sequence advances.” Then test it against “Regional partition owner becomes unavailable”. State the retry identity, durable checkpoint, recovery signal, and cost of the alternative.

Answer pattern: state the invariant, choose a mechanism, walk one failure, name the recovery signal, and explain the cost of the alternative.

↑ top

11. How to present this in 45 minutes

Time	What to cover
0–5 min	Clarify scope, workload, SLOs, and the correctness invariant.
5–10 min	Estimate throughput/storage and define concrete APIs and records.
10–20 min	Draw online, durable-write, and asynchronous/repair paths.
20–30 min	Explain the exact lifecycle, partition key, and concurrency control.
30–38 min	Discuss caching, hot keys, replication, and backpressure.
38–43 min	Run failure drills, security, observability, and regional evolution.
43–45 min	Summarize the five most important choices and trade-offs.

Final decision summary

Invariant	accepted events remain ordered within their conversation/session and survive reconnects.
Partitioning	Stable aggregate/key hash with versioned ownership and measured hot-key isolation.
Commit	Expected version or uniqueness at the authoritative state boundary.
Async work	Durable outbox/log, idempotent consumers, checkpoints, and replay.
Recovery	Same operation identity, durable truth lookup, and business reconciliation.

Closing summary

CLOSING “The design protects accepted events remain ordered within their conversation/session and survive reconnects. The online path stays bounded, authoritative state is versioned and partitioned, and asynchronous work is replayable and observable.”

↑ top

Mobile / Web

Anycast Edge

WebSocket Gateway

Session Directory

Complete System Design Tutorial

1. Requirements and scope

Functional requirements

Non-functional requirements

2. Capacity assumptions

3. API and data model

Authoritative records and indexes

Command and event envelopes

4. Why this architecture and what it costs

Chosen boundary vs simpler monolith

Correctness vs availability

Operational cost and evolution

5. Critical flow and algorithm

Retry-safe mutation skeleton

6. Caching, hot keys, and backpressure

7. Partitioning and consistency

Strong boundary

Eventual boundary

8. Failure drills and recovery

Recovery checkpoint

9. Security, regions, and observability

Security and abuse

Signals that prove correctness

Regional evolution

10. Interviewer follow-ups

11. How to present this in 45 minutes

Final decision summary

Closing summary