How would you design the data structures for a social network (Social Graph)?

Step 1: Use Cases, Constraints, Assumptions, Back-of-Envelope

Use cases

We'll scope to handle only the core flows that expose the bottlenecks; everything else is deferred.

• User adds/removes friend — mutual edge write, idempotent, replicated with quorum.
• User reads friends / friends-of-friends — adjacency lookup, paginated, filtered by activity.
• User reads feed derived from graph — 2-hop traversal (friends' posts) with cache.
• Service suggests friends — FoF (friend-of-friend) ranking via mutual count.

Out of scope

• Messaging, media posts themselves, auth, privacy ACL beyond edge visibility (separate policy engine).

State assumptions

• 500M users, avg 200 friends → 100B edges (undirected stored as 2 directed =200B rows).
• Edge 16 bytes (28) + overhead 32B ≈ 48B → 200B48≈9.6 TB raw, 30 TB with replicas+indexes.
• Read QPS: friend-list 10k/s, FoF