Design a scalable tagging system
Overview
This question evaluates understanding of distributed systems, data modeling, indexing, sharding, caching, consistency models, and multi-tenant architecture for building high-throughput tagging and query services.
Atlassian
Aug 4, 2025, 10:55 AM
Software Engineer
Technical Screen
System Design
0
0
System Design: Scalable Tagging Service
Context
You are designing a multi-tenant tagging service for a large-scale SaaS product. Items (e.g., documents, issues, pages) can have multiple tags. The system must support high read/write throughput and efficient queries by one or multiple tags.
Assume scale on the order of:
- 100M items, 10M distinct tags, average 5–20 tags per item
- Peak 10k read QPS, 2k write QPS (tag add/remove)
- Multi-tenant isolation is required
Requirements
- Create and manage tags
- Associate/dissociate multiple tags with an item
-
Query items:
- by a single tag
- by a combination of tags (AND/OR), with pagination and sort (e.g., recency)
- High availability and low latency for reads; scalable writes
- Consistency model: clearly define and implement (eventual vs read-your-writes)
- Sharding, caching, and indexing strategies for scale
- Operability: monitoring, reindexing, backfills
Deliverables
- Data model and storage choices
- Read and write paths (including APIs)
- Indexing strategy for tag queries (single and combined)
- Sharding/partitioning plan
- Caching strategy and invalidation
- Consistency trade-offs and mechanisms
- Handling hot tags, skew, and growth
Solution
Login required
Show