Pilot Light
A disaster recovery pattern where only the core data layer runs in a second location, with everything else switched on and scaled up on failover.
Definition
Pilot Light is one of the four standard AWS disaster recovery patterns. The core of a system — usually its database — is continuously replicated to a second location, while the rest of the infrastructure is kept switched off or minimally provisioned. On failover, that infrastructure is started and scaled up around the data that is already there. Recovery typically takes tens of minutes to a few hours, and data loss is usually measured in minutes, at a much lower running cost than keeping a full copy live. It sits between Backup & Restore (cheaper, slower) and Warm Standby (faster, more expensive).
In Practice
A payments reconciliation system that can tolerate an hour of downtime but not a day of lost data is a good fit for Pilot Light: its database replicates continuously, and its application servers are only launched when a failover is declared.