Actual availability is a metric that reflects the real amount of time an IT system, service, piece of equipment, or infrastructure remains available over a specified period. It shows the proportion of time the system was actually operational and accessible to users, as opposed to a target or guaranteed availability level defined, for example, in an SLA.
Actual availability is used to assess the reliability of cloud services, servers, networks, data centers, applications, and other IT systems. It is usually expressed as a percentage: for example, 99.9% availability means that the system was available for 99.9% of the measured time during the period under review.
How Is Actual Availability Calculated?
A basic calculation uses the following formula:
Actual availability = (total time − downtime) / total time × 100%
For example, if a service was unavailable for a total of 60 minutes during a 30-day month, its actual availability would be approximately 99.86%.
However, accurately calculating availability requires more than simply recording the duration of all outages. The measurement methodology must be defined in advance. In particular, it should specify:
- what qualifies as service unavailability
- where monitoring is performed from and how frequently checks are made
- whether scheduled maintenance is included
- whether outages caused by third-party carriers or providers are counted
- whether incidents caused by the customer are excluded
As a result, the same number of downtime minutes does not always produce the same officially calculated availability. The final figure depends on the rules used to record and classify incidents.
Actual Availability and SLA
An SLA (Service Level Agreement) may define a guaranteed service availability level, such as 99.9%, 99.95%, or 99.99%. Actual availability represents the result that was actually achieved during a specific reporting period.
For example, if the SLA requires availability of at least 99.9% and the actual availability for the month is 99.97%, the requirement has been met. If the measured value falls below the specified threshold, the agreement may provide for service credits or other forms of compensation.
Therefore, SLA availability and actual availability are different metrics:
- SLA availability defines the provider’s contractual commitment
- actual availability reflects the system’s measured operational performance
When comparing the two, it is important to use the same calculation methodology. If an SLA excludes agreed maintenance windows while an external monitoring system counts them as downtime, the resulting availability figures may differ.
What Affects Actual Availability?
The actual level of availability depends on more than the reliability of individual servers. Modern IT services typically consist of multiple interconnected components, so the failure of any critical element can affect overall service operation. Key factors include:
- redundancy of servers, storage systems, and network equipment
- availability of independent network connections and data transmission paths
- redundant power supplies and supporting infrastructure
- fault-tolerant application and database architecture
- software quality and update procedures
- speed of incident detection and resolution
High availability is therefore generally achieved not through a single exceptionally reliable component, but through an architecture that allows the system to continue operating when individual components fail.
How Is Actual Availability Measured?
Actual availability is measured using monitoring systems that regularly check service status and record the beginning and end of periods of unavailability. Monitoring may check a website’s HTTP response, the availability of a specific port, an API, a network node, an application, or another monitored component.
In distributed systems, monitoring is often performed from multiple geographic locations. This helps distinguish a global service outage from a local routing problem or an accessibility issue affecting a particular region.
The availability percentage alone also does not provide a complete picture of reliability. For example, ten six-minute outages and a single one-hour outage result in the same total downtime but may affect users and business processes very differently. For this reason, actual availability is typically analyzed together with the number, duration, and causes of incidents.