Skip to main content
Legia

IT Infrastructure & Cloud

What Is Uptime and How Is It Calculated?

Uptime is the percentage of time a server or service works within a period. Here is how to calculate it, what small differences in the percentage mean, and what to agree on before the figure is used.

SEO and Content Specialist

2 min read

Yellow fiber optic cables on a green connector panel

Short answer: uptime is the percentage of time a server or service works within a period. The formula: (total minutes in the period minus minutes of downtime) divided by total minutes in the period, times 100. The figure only means something when the period, the definition of “down”, the treatment of scheduled maintenance, and the measuring tool are written down clearly.

Uptime appears in almost every hosting, cloud, and server management offer. How to calculate server uptime is simple. What is often disputed is what counts as downtime. How to choose metrics and alert thresholds is covered in the parent article Server Monitoring: Metrics and Alert Thresholds. Here the focus is on the calculation and how to read the figure.

How to calculate server uptime

Take one period, for example a 30-day month: 30 x 24 x 60 = 43,200 minutes. If the server was unreachable for 90 minutes that month:

(43,200 - 90) / 43,200 x 100 = about 99.79%

A small difference in percentage means a large difference in downtime. The tables in the Google SRE Book and the Microsoft Azure Well-Architected Framework carry the same figures:

UptimeDowntime still within target per monthPer year
99%7.2 hours3.65 days
99.9%43.2 minutes8.76 hours
99.99%4.32 minutes52.56 minutes

So the 99.79% in the example above does not meet a 99.9% target, even though the numbers look close.

Illustration: server

Four things to agree on before using the figure

  1. Period. Monthly, quarterly, or yearly gives different figures for the same events.
  2. Definition of “down”. The server cannot be reached, the application returns errors, or it is extremely slow? Request-based measurement, as Google uses, counts the ratio of successful requests, not just whether the server is on.
  3. Scheduled maintenance. Counted as downtime or excluded? The Google SRE Book table assumes no planned downtime.
  4. Measuring tool. A figure you can check comes from recorded monitoring, not from a one-sided report.

Uptime of several components

When a service depends on several components in series, the combined percentage is lower than each component. The Azure Well-Architected Framework describes the composite value as the product of each factor’s contribution. A worked example: a server at 99.95% and an internet connection at 99.9% give a combined figure of about 99.85%.

In Legia’s services

In Server Installation & Management, the availability level requested is one of the bases for the cost calculation, and automatic monitoring sends early warnings as soon as a problem appears. The monthly report contains server availability and incidents, so the uptime figure can be checked against the records. Other terms in the contract, such as response time, are explained in What Is an SLA in an IT Service Contract?.

Sources

Category IT Infrastructure & Cloud

Keep reading

Related articles

All Articles