Preparing your design workspace

Loading the problem and your latest saved attempt…

Offline Multi-Device Data Sync

Design a client data-editing system that remains usable offline and reconciles changes across devices.

Problem statement

For this exercise, assume the source-backed system identity and the explicitly bounded practice contract that follow. Design the client and synchronization architecture for an application that edits structured user data.

Each device must keep a durable local copy, accept edits while disconnected, synchronize after connectivity returns, and resolve concurrent changes made on multiple devices.

For this exercise, assume record deletion, device expiry, arbitrary binary files, cross-account collaboration, and real-time coediting are out of scope. The structured record model and conflict policy remain open interview inputs; distinguish durable user intent from caches and derived projections.

What to build

  1. Read and edit a durable local working set without network access, including across an application restart.
  2. Exchange local and remote changes after reconnect without silently losing an acknowledged edit.
  3. Detect concurrent multi-device changes and apply an explicit deterministic merge or user-visible conflict outcome.
  4. Expose pending, synchronized, conflicted, and failed states so a user can understand whether an edit is durable and replicated.

At scaleExercise boundary

  • Support up to 100,000 locally relevant structured records per user.
  • Support up to four active devices per user; each device may remain offline for seven days and accumulate up to 10,000 pending changes before reconnecting.
  • A local edit must become durable on its originating device within 50 ms at p95.
  • After stable connectivity returns, background synchronization should converge within 30 seconds at p95 for an ordinary reconnect batch.