Doc Collaboration

Doc Collaboration with personalized feeds, graph edges, and engagement.

Requirements

Functional Requirements

• Publish posts
• Render personalized timelines
• Record engagement

Non-functional Requirements

• Low p95 read latency
• Freshness within seconds

High-Level Design

• Write → fanout → timeline store → rank

Capacity & Sizing

• Posts/sec, followers distribution, timeline storage

Key Components

• Write service, Fanout, Ranking, Timeline store

Architecture

Data Model

• posts (post_id PK, author_id, ts, body)
• edges (follower_id, followee_id)
• timeline (user_id, post_id, score, ts)

APIs

• POST /api/posts { body }
• GET /api/timeline?user=..
• POST /api/posts/:id/like

Hot Path

1. Read → candidate select → rank → cache

Caching & TTL

• Cache timelines for seconds; invalidate on new posts

Scaling

• Hybrid fanout by follower count
• Cache user timelines
• Async recompute

Trade-offs

• Freshness vs cost
• Staleness windows
• Cold start

Failure Modes & Mitigations

• Ranking outage → recency
• Cache storms → jittered refresh
• Mega influencers → isolation

Observability

• p95 read latency
• cache hit ratio
• fanout lag