Google Docs#

Q: How does Google Docs work in real time?

Every keystroke is sent as an operation over WebSocket to a collab server. The server transforms concurrent ops with OT and broadcasts the result to all clients, keeping every replica converged.

Q: What is Operational Transformation in Google Docs?

OT rewrites concurrent ops so they apply in the same order on every replica. If Alice inserts at index 3 and Bob inserts at index 5 simultaneously, the server adjusts indices so both edits land correctly.

Q: OT vs CRDT for Google Docs?

Google chose OT because it produces compact ops and a central server scales well for the editor's traffic. CRDTs trade larger payloads for peer-to-peer and offline-first merges.

Q: How does Google Docs handle offline edits?

Offline ops are buffered locally with a base revision tag. On reconnect they replay against the server, which transforms each against any newer ops before applying them.

Q: How is Google Docs revision history stored?

The collab server appends every op to an ordered log per document. Revision history reconstructs any past state by replaying ops up to a target version.

Problem statement (interviewer prompt)

Design Google Docs: many users edit the same document simultaneously and see each others' edits + cursors in real time. Pick between Operational Transform (OT) and CRDTs, handle offline edits + sync, comments + suggestions, and revision history.

flowchart LR
  A([User A])
  B([User B])
  WS[WebSocket Gateway]
  COL[Collab Service<br/>OT / CRDT]
  DOC[(Doc state + history)]
  A --> WS --> COL --> DOC
  B --> WS
  COL --> WS --> B

    classDef client fill:#dbeafe,stroke:#1e40af,stroke-width:1px,color:#0f172a;
    classDef edge fill:#cffafe,stroke:#0e7490,stroke-width:1px,color:#0f172a;
    classDef service fill:#fef3c7,stroke:#92400e,stroke-width:1px,color:#0f172a;
    classDef datastore fill:#fee2e2,stroke:#991b1b,stroke-width:1px,color:#0f172a;
    classDef cache fill:#fed7aa,stroke:#9a3412,stroke-width:1px,color:#0f172a;
    classDef queue fill:#ede9fe,stroke:#5b21b6,stroke-width:1px,color:#0f172a;
    classDef compute fill:#d1fae5,stroke:#065f46,stroke-width:1px,color:#0f172a;
    classDef storage fill:#e5e7eb,stroke:#374151,stroke-width:1px,color:#0f172a;
    classDef external fill:#fce7f3,stroke:#9d174d,stroke-width:1px,color:#0f172a;
    classDef obs fill:#f3e8ff,stroke:#6b21a8,stroke-width:1px,color:#0f172a;
    class A,B client;
    class WS edge;
    class COL service;
    class DOC datastore;

flowchart TB
  subgraph Clients
    EDIT[Editor app]
    PRES[Presence]
    CUR[Cursor / selection]
  end

  subgraph Edge
    LB
    WS[WebSocket gateway<br/>pinned per doc]
  end

  subgraph Collab[Collaboration engine]
    OT[OT engine<br/>transform op against concurrent ops]
    CRDT_ENG[CRDT engine<br/>Yjs / Automerge style]
    SESS[Session manager]
    REV[Revision log per doc]
    SNAP[Snapshots]
    LOCK[Per-doc serialization]
  end

  subgraph Storage
    DOC[(Doc store)]
    OPLOG[(Op log)]
    META[(Permissions, comments, suggestions)]
  end

  subgraph Features
    PRES_S[Presence service]
    COMM[Comments / Threads]
    SUGG[Suggesting mode]
    VER[Version history]
    EXPORT[Export PDF / Word]
    OFF[Offline support]
  end

  subgraph Share
    ACL[ACL / link sharing]
    AUTH[Auth + tenant]
  end

  Clients --> LB --> WS --> Collab
  Collab --> Storage
  Features --- Collab
  Share --- Storage

    classDef client fill:#dbeafe,stroke:#1e40af,stroke-width:1px,color:#0f172a;
    classDef edge fill:#cffafe,stroke:#0e7490,stroke-width:1px,color:#0f172a;
    classDef service fill:#fef3c7,stroke:#92400e,stroke-width:1px,color:#0f172a;
    classDef datastore fill:#fee2e2,stroke:#991b1b,stroke-width:1px,color:#0f172a;
    classDef cache fill:#fed7aa,stroke:#9a3412,stroke-width:1px,color:#0f172a;
    classDef queue fill:#ede9fe,stroke:#5b21b6,stroke-width:1px,color:#0f172a;
    classDef compute fill:#d1fae5,stroke:#065f46,stroke-width:1px,color:#0f172a;
    classDef storage fill:#e5e7eb,stroke:#374151,stroke-width:1px,color:#0f172a;
    classDef external fill:#fce7f3,stroke:#9d174d,stroke-width:1px,color:#0f172a;
    classDef obs fill:#f3e8ff,stroke:#6b21a8,stroke-width:1px,color:#0f172a;
    class WS edge;
    class EDIT,PRES,CUR,OT,CRDT_ENG,SESS,REV,SNAP,LOCK,PRES_S,COMM,SUGG,VER,EXPORT,OFF,ACL,AUTH service;
    class DOC,OPLOG,META datastore;

OT vs CRDT#

Operational Transform (Google Docs historically): server orders ops and transforms incoming ops against concurrent committed ops.
CRDT (Yjs, Automerge): each op has unique ID with causal context; merges by mathematical guarantees.

Per-doc serializer#

One server owns a doc at a time → serializes ops in a definite order.
Failover to another server reloads from op log.

Presence#

Lightweight; per-doc pub/sub channel.
Cursors and selections sent at 10 Hz max with throttling.

Glossary & fundamentals#

Concepts referenced in this design. Each row links to its canonical page; the tag column shows whether it is a high-level (HLD) or low-level (LLD) concept.

Tag	Concept	What it is	Page
`HLD`	Pub/Sub & message brokers	topics, consumer groups, delivery semantics	pub-sub-pattern
`HLD`	CRDTs	commutative replicated data types	crdts
`HLD`	Realtime protocols	WS / SSE / polling / gRPC streaming	realtime-protocols
`LLD`	Async models	futures / async-await / coroutines / actors	async-models

Quick reference#

Functional#

Concurrent editing with sub-second sync.
Cursor + selection presence.
Comments, suggesting, version history.
Offline mode + reconciliation.
Sharing, link tokens, ACLs.

Non-functional#

p99 op echo < 200 ms.
100+ concurrent editors per doc.
99.99% uptime.

Trade-offs#

OT vs CRDT: OT mature with central server; CRDT decentralizable + offline-friendly.
Per-doc serialization simplifies correctness; failover challenge.
Storage: op log → periodic snapshots to bound replay cost.

Refs#

Google Wave / Docs OT papers.
Yjs / Automerge / RGA papers.
ByteByteGo "Design Google Docs".

FAQ#

How does Google Docs work in real time?#

Every keystroke is sent as an operation over WebSocket to a collab server. The server transforms concurrent ops with OT and broadcasts the result to all clients, keeping every replica converged.

What is Operational Transformation in Google Docs?#