ByteDance System Design: Reward Backend for High QPS

Overview

You are asked to design a scalable, resilient backend for a mobile app rewarding mechanism where users earn and redeem points. The system must sustain high QPS (e.g., 10k req/s) during promotions while preventing double-spend, preserving auditability, and keeping latency low.

Core problem & constraints

You should cover accurate crediting and deduction of points, near-real-time balance updates, audit logging, user notifications, and graceful behavior under load spikes. Non-functional constraints include 99.9% availability, sub-200ms latency for most requests, fault tolerance, and cost-efficiency. Expect trade-offs between strong and eventual consistency for throughput.

High-level flow / interview stages

• Ingest: API gateway + rate limiter + authentication
• Validation: verify action (ad view, purchase, task) possibly via event callbacks
• Accounting: update user balance (fast path via cache/sharded counter, durable write to ledger)
• Redemption: reserve points, atomically complete deduction, trigger fulfillment
• Audit & monitoring: append immutable transaction logs (event stream) and write to OLTP for reconciliation

In an interview you'll be expected to sketch components, explain data flow, and justify consistency and failure-handling choices.

Skill signals interviewers look for

• Designing stateless services, load balancing, and autoscaling
• Caching strategies (read-through/write-through, sharded counters) and cache invalidation
• Rate limiting (token bucket, per-user/global tiers) and backpressure
• Concurrency control and idempotency to avoid double-spend (optimistic locks, distributed locks, compare-and-swap)
• Durable event logging (Kafka/event sourcing) and reconcilers for eventual consistency
• Monitoring, SLA-aware retries, circuit breakers, and cost/performance trade-offs

Prepare to discuss concrete technologies (Redis, Kafka, RDBMS), sharding, and fallback behavior during outages.