← Lessons

quiz vs the machine

Gold1450

System Design

Active Active Topology

Running every node hot so capacity is used and failover is instant.

5 min read · core · beat Gold to climb

The idea

In an active active topology all nodes serve traffic at the same time. There is no idle standby; every replica does useful work and shares the load.

The payoff is twofold: you use all the capacity you pay for, and losing one node simply shifts its share to the others with no promotion step.

What makes it hard

The challenge is state. If two active nodes can both write the same data, you must answer how conflicts are resolved.

  • Stateless services are easy to run active active because requests carry no node specific state.
  • Stateful stores need either partitioning so each key has one owner, or conflict resolution like last writer wins or merge functions.

Capacity planning

A subtle trap is sizing. If two nodes each run at seventy percent and one fails, the survivor needs to absorb the load and will be overloaded. Active active fleets must run with headroom so survivors can carry the extra share.

Routing

Load balancers spread requests across all nodes and stop sending to any that fail health checks, so recovery is automatic and fast.

Key idea

Active active uses all capacity and fails over instantly, but only if state conflicts are handled and headroom is reserved.

Check yourself

Answer to earn rating on the learn ladder.

1. Why must active active fleets keep headroom?

2. Which is easiest to run active active?