| Users | Transactions/sec | Database Load | System Changes |
|---|---|---|---|
| 100 users | 10-50 | Single DB instance handles writes and reads | Simple ACID transactions, no sharding needed |
| 10,000 users | 1,000-5,000 | DB CPU and I/O increase, some read replicas added | Introduce connection pooling, caching for reads |
| 1,000,000 users | 50,000-100,000 | Single DB cannot handle writes, bottleneck at DB | Sharding writes, distributed transactions, async processing |
| 100,000,000 users | 5,000,000+ | Network and storage bottlenecks, complex consistency | Microservices, event sourcing, CQRS, global data centers |
Why e-commerce tests transactional design in HLD - Scalability Evidence
In e-commerce transactional design, the first bottleneck is the database write capacity. As users increase, the number of transactions (orders, payments, inventory updates) grows. A single database instance can only handle a limited number of writes per second (usually a few thousand). This causes delays and potential data inconsistency if not managed properly.
- Read Replicas: Offload read queries to replicas to reduce load on primary DB.
- Connection Pooling: Efficiently reuse DB connections to handle more concurrent requests.
- Caching: Use in-memory caches (e.g., Redis) for frequently read data like product info.
- Sharding: Split database by user or order ID to distribute write load across multiple DBs.
- Asynchronous Processing: Use message queues to handle non-critical updates later, reducing immediate DB load.
- Eventual Consistency: Relax strict ACID for some operations to improve scalability.
- Microservices: Separate transactional domains (orders, payments, inventory) for independent scaling.
Assuming 1 million users with 100,000 transactions per second:
- Database writes: 100,000 QPS (queries per second) - requires multiple DB shards.
- Storage: Each transaction ~1 KB, daily storage ~8.6 TB (100,000 tx/sec * 1 KB * 86400 sec).
- Network bandwidth: 100,000 tx/sec * 1 KB = ~100 MB/s sustained write bandwidth.
- Cache: To reduce DB reads, cache size ~100 GB for hot product data.
- Servers: Multiple app servers (100+) to handle concurrent users and transactions.
When discussing e-commerce transactional design scalability, start by identifying the transaction volume and database write limits. Explain how ACID properties affect scaling. Then, describe solutions like sharding and asynchronous processing. Always mention trade-offs between consistency and availability. Use real numbers to show understanding.
Your database handles 1,000 QPS. Traffic grows 10x to 10,000 QPS. What do you do first and why?
Answer: The first step is to add read replicas and implement connection pooling to reduce load on the primary database. For writes, consider sharding or asynchronous processing to distribute and delay load. This prevents the DB from becoming a bottleneck and maintains transaction integrity.
