System Design: Historical FX Rates Service (10k QPS)

Background

You are designing an internal service for engineers and analysts to fetch historical currency exchange rates (e.g., USD→EUR) for analytics, backfills, and financial reporting. The service must be highly available, low latency, and cost-efficient, and will be consumed by internal applications and data pipelines.

Assume the service primarily supports read traffic (10k QPS) with occasional corrections/backfills from upstream providers. Historical rates are mostly immutable but can be revised.

Requirements

Describe a design that covers:

1. APIs
   • Read endpoints for point-in-time and time-series queries.
   • Optional write/ingestion endpoints for internal data pipelines.
   • Versioning, pagination, auth, rate limits, and error handling.
2. Data Model
   • Entities, fields (e.g., pair, timestamp, rate types), precision, versioning for corrections, and indexing.
3. Storage Layer
   • Hot vs. cold storage choices; partitioning/sharding; replication; retention.
4. Caching Strategy
   • Layers (in-process, distributed, CDN if applicable), keys, TTLs, invalidation, and precomputation.
5. Consistency Requirements
   • Freshness expectations, handling of corrections/backfills, snapshot semantics, and trade-offs (strong vs. eventual consistency).
6. Scalability Plan
   • Capacity estimates for 10k QPS, autoscaling, multi-region strategy, and failure handling.
7. Monitoring and Operations
   • SLIs/SLOs, alerts, tracing, data quality checks, and runbooks.

State reasonable assumptions where needed to make the problem concrete.