Configuring for higher availability or resilience Is like buying insurance

In selecting insurance policies, you typically weigh the cost of the insurance against the likelihood that the insurance is needed, whether the insurance is required by law, and the significance of the potential loss if you don't have insurance.

For example, you would likely not take a flood insurance policy regardless of the premium cost if you live on a hilltop in a desert but you would buy a high premium flood insurance if you live in a hurricane zone along the coast. Similarly, when procuring system hardware, you buy servers with high reliability, availability and serviceability (RAS) built in to ensure that hardware faults do not result in an outage. For example, your servers may come equipped with as many as six redundant cooling fans and power supplies.

In some cases, the law may require you to purchase insurance. Similarly, in some business sectors, regulations require business continuity and disaster recovery plans.

At the extremes, if a business is willing and able to tolerate prolonged outage periods, the HA requirements are few. In some cases, having good backups may suffice.

On the other hand, if a business can only tolerate a down time of less than 30 minutes for each outage, you may have to consider having duplicated or redundant components for any component that can fail especially if they are the SPOF.

At the other end of the spectrum, a company may have very high availability requirements and can only tolerate less than five minute downtime for each outage. In that environment, the data center may have to be staffed around the clock, the failure detection must be quick, failover procedures must the automated, and so forth.