Fair Multitenancy: Rate Limiting for SaaS APIs

Fair multitenancy goes beyond simple rate limiting to ensure each SaaS tenant gets a fair share of infrastructure. Key distinctions include rate vs. concurrency limits, scoping limits by entity/action/region, and defining tiered or adaptive quotas per customer. Concurrency limits are harder to implement in distributed systems due to horizontal scaling challenges. Good practices include documenting limits publicly, providing human-friendly error messages, maintaining a centralized limit store with override capabilities, using leaky buckets to smooth bursts, and running simulation modes before enforcing limits. The post also cautions against overly aggressive rejection logic and suggests single-tenant deployments for very large customers.