A cloud application rarely runs on a single machine. Instead, a load balancer sits in front of a pool of servers spread across several availability zones — independent data centres with their own power and networking. This scene shows one cloud "region" from above: a load balancer beacon routes incoming request packets to healthy, auto-scaling server pools, and returns the response to the client.
Major cloud providers guarantee "N+1" or better redundancy for critical services — meaning the system keeps running even if one entire availability zone fails. That guarantee is a core reason cloud architectures favour many small, replaceable instances over one large server.
A cloud region rendered in 3D: a load balancer routes request packets to auto-scaling server pools spread across three availability zones, showing how scalability, redundancy and cost trade off against each other.
Auto-scaling adds and removes server instances to match traffic, multiple availability zones provide failover when one goes down, and every running instance carries a real monthly cost — the three pillars of cloud architecture design.
Raise the traffic load and watch instances scale up (or overload, if auto-scaling is off). Reduce deployed zones to see the cost drop, then simulate an outage to see what redundancy actually buys you.
Cloud providers price availability zones and always-on redundancy as a trade-off: doubling your zones roughly doubles your baseline infrastructure cost, which is why architects size redundancy to the actual cost of downtime.