- Backup-and-Restore
- The frugal end of the recovery spectrum: data is copied periodically to a separate location, and recovery means provisioning new infrastructure and restoring onto it after a failure. Slow to recover and bounded in data loss by backup frequency, but nearly free while systems are healthy — the honest answer for systems that can genuinely tolerate long outages.
- Business Impact Analysis
- The survey conducted before recovery investment that establishes, for each significant process, what its failure costs, how fast that cost accumulates, and what it depends on to run. Its most valuable output is often a hidden load-bearing dependency rather than a cost figure. It is perishable and must be maintained as the organisation changes.
- Recovery Point Objective (RPO)
- The maximum acceptable amount of recent data that may be lost in a failure, expressed as a window of time reaching backward from the moment of failure. A measure of tolerable forgetting, governed by how often and how continuously data is copied away from the system that might fail. Independent of RTO.
- Recovery Time Objective (RTO)
- The maximum acceptable length of time a system or process may be unavailable following a failure before the resulting harm becomes unacceptable. A measure of the organisation's patience, set by business owners and translated into architecture by engineers. Distinct from the recovery time actually achieved, which only testing reveals.
- Shared Fate
- The condition in which a system and its recovery site depend on something in common — power, network path, location, account, control plane, deployment pipeline, or an external provider — so that a single failure can take both at once. The chief way independent-looking designs fail together; found by relentlessly asking what a backup shares with the thing it protects.
- Testing Ladder
- The progression of recovery tests from cheap and safe to expensive and real: the tabletop exercise (discussion, finding missing decisions), the recovery drill (executing recovery in isolation, measuring real recovery time and exposing drift), and the live failover (failing real production over under genuine load). Each rung finds failures the rungs below it structurally cannot.
- Tiering
- The practice of sorting systems into a small number of bands, each with its own recovery objectives and budget, so protection matches harm. Tier assignment follows business consequence tempered by dependency, and tiers must carry genuinely different architecture and cost or the exercise is theatre.
- Warm Standby
- A scaled-down but fully functional copy of a system running continuously alongside production, ready to take over and scale up within minutes of a failure. A pragmatic middle-to-high point on the spectrum that meets demanding recovery times without the full cost of running two complete production environments.