Cell-Based Architecture Versus Shared Multi-Tenant Stack
TradeoffPrincipalUpdated
Your B2B SaaS runs 4,000 tenants on one shared stack, and twice this year a bad deploy or one noisy tenant took every customer down. Enterprise contracts promise 99.95% availability.
Short answer: Cell-Based Architecture. Split the fleet into cells, each a full copy of the stack serving a fixed slice of tenants, so any failure reaches only one cell.
The Two Options
A
Cell-Based Architecture
B
Shared Multi-Tenant Stack
Drill This: Your Pick First
The Better Pick
Option A
Cell-Based Architecture
Split the fleet into cells, each a full copy of the stack serving a fixed slice of tenants, so any failure reaches only one cell. Deploys roll out cell by cell, and a noisy tenant can exhaust only its own cell's capacity. The cost is a routing layer, more fleets to operate, and tooling to move tenants between cells.
When the Other One Wins
Shared Multi-Tenant Stack
When outages have been rare and tenants are small, a shared stack with per-tenant rate limits and staged deploys gives most of the protection at a fraction of the operating cost.