Bytedance·Software Engineer·Onsite - System Design / Architecture·Senior

SeniorPrefer not to say

Jun 2026

Summary

Bytedance system design round, one big question about building a flash-sale coupon service at global scale. Dense question with a lot of moving parts and I'm not sure I covered everything they wanted.

Questions Asked (1)

Design a global flash-sale coupon-grab system where users across many countries compete for a limited number of coupons in a short time window. The system must guarantee equal opportunity regardless of network latency or geography, handle massive traffic bursts, and never over-claim inventory. Walk through your data store choices, regional architecture, how you'd split coupon inventory across regions, queueing strategy, and rate limiting.

System DesignTechnical Trade-offsData Modeling

Author's notes

This question is deceptively wide.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by clarifying requirements and constraints, then propose a globally distributed architecture that ensures fairness through synchronized start times and regional inventory partitioning. Focus on strong consistency for inventory deduction using atomic operations in a fast data store, and use queues and rate limiting to handle bursts and prevent over-claiming.

Pro tip: Emphasize that perfect fairness is impossible due to physics, but you can achieve perceived fairness by synchronizing clocks, using a lottery-based approach for excess demand, and ensuring transparency in the process.

1. Clarify Requirements and Constraints

Ask about scale (users, coupons), latency requirements, consistency vs. availability trade-offs, and whether exact fairness is required or perceived fairness is acceptable.

2. Design Regional Architecture

Deploy the system in multiple regions with edge servers to reduce latency. Use a global load balancer to route users to the nearest region, and synchronize the sale start time across all regions using a reliable time service.

3. Choose Data Stores and Inventory Partitioning

Use a fast, in-memory data store like Redis with atomic decrement operations for inventory. Partition coupon inventory across regions proportionally to user base or expected demand, and use a global coordinator to rebalance if needed.

4. Implement Queueing and Rate Limiting

Place requests into a regional queue (e.g., Kafka) to absorb bursts. Apply rate limiting per user and per region to prevent abuse and ensure fair access. Process queue with a controlled rate to match inventory deduction capacity.

5. Ensure Consistency and Prevent Over-claiming

Use atomic operations (e.g., Redis DECR) or distributed locks to guarantee no over-claiming. Implement idempotency to handle retries, and use a two-phase commit or saga pattern for cross-region inventory adjustments if needed.

Key Points to Mention

Use of Redis or similar in-memory store with atomic operations for inventory deduction.
Regional partitioning of inventory to reduce latency and contention, with a global coordinator for fairness.
Synchronized start time using NTP or a global clock service to ensure equal opportunity.
Queueing (e.g., Kafka) to handle traffic bursts and smooth out spikes.
Rate limiting per user and per IP to prevent abuse and ensure fair access.
Idempotency and retry mechanisms to handle duplicate requests and network failures.

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.