A design property of a system that aims for very low downtime, typically 99.9%+, by removing single points of failure.
High availability (HA) is the design discipline of building systems with very low downtime targets, typically 99.9% (~43 min/month) or better, by eliminating single points of failure. HA architectures use replication, failover, multi-AZ or multi-region deployments, redundant load balancers, automated leader election, and health-checked routing. The hardest part is not the architecture, it's the operational discipline of testing the failover paths and keeping them current as the system evolves.
HA targets compound: 99.9% (3 nines) is 8.76 hours of downtime per year; 99.99% (4 nines) is 52 minutes. The cost to move between tiers is roughly geometric: each extra nine is roughly 10x the engineering effort and infrastructure cost. Picking the right target for the right service tier (customer-facing checkout = 4 nines, internal admin tool = 3 nines) is the most impactful early architecture decision.
See the part of the platform that handles high availability (ha) in production.