Cell-based multi-tenant routing

DesignHard45 minAmazon and Salesforce style

Your task

Design a cell-based (sharded-by-tenant) architecture for a SaaS platform serving 1M tenants, where the goals are: limit the blast radius of any failure to a small fraction of tenants, route every request to the correct cell with minimal latency, and migrate a tenant from one cell to another (rebalancing, or moving a growing tenant to a dedicated cell) with near-zero downtime. Walk through how tenant to cell mapping works and is looked up, how you keep the router fast and correct, and how a live tenant migration avoids split-brain.

It covers

  • Tenant routing
  • Cell architecture
  • Blast radius