Visit booth 3171 at Google Cloud Next to see how to unlock real-time decisions at scaleMore info
Glossary

What is five-nines uptime?

At a glance:

  • Five-nines uptime: A common shorthand for a device, network, or service that's available to users 99.999% of the time in a specified period (usually one year).

  • Also called: 5 nines uptime, five 9s uptime, 99.999% uptime, high availability, always on

  • Availability: The availability of a service measures the percent of time when a user can reach that service. Availability uses historical performance to estimate future performance. 

  • What is the purpose of five-nines? Buyers use availability to evaluate vendors for reliability, while service providers use it to establish performance goals and find opportunities for improvement.

  • What is five-nines of downtime per year? Five-nines systems experience 5.26 minutes of downtime in a year, or as long as two TV commercial breaks.

  • Related concepts: System uptime, fault-tolerant system, service level agreement (SLA), service level objective (SLO)

In IT, uptime is often measured in "nines": how many nines of availability does your service provide, from two-nines (99% uptime) to five-nines (99.999% uptime), in a given period of time. Five-nines uptime translates to 5.26 minutes of downtime per year. 

Five 9s uptime can be included as a guarantee in an SLA, an internal goal in an SLO, or as predicted availability based on past service. Five-nines is often considered "always on."

What is five-nines of availability?

Adding another nine after the decimal might seem like a small change, but it's a big difference: the delta between two nines and fine nines availability is 5,250.74 minutes — more than 3.5 days.

Why is fine nines uptime important?

If you operated an emergency service and every second put lives in jeopardy, how much downtime would be acceptable?

Most businesses don't measure uptime in lives, but the stakes are still astronomical. The Crowdstrike-Microsoft outage on July 19, 2024, is estimated to have resulted in $5.4 billion in direct costs to Fortune 500 companies, excluding Microsoft. It lasted 79 minutes, or 99.985% uptime (assuming no other outages). 

The high cost of disruptions is why many consider five-nines the minimum standard for acceptable availability.

Measuring five-nines availability

Availability is calculated as follows:

Uptime (Period - Downtime) / Period

The real world is rarely that simple. Three variables can affect uptime:

  1. What counts as downtime:

Not all downtime is counted. Service providers usually include:

  • Scheduled downtime for maintenance

  • Unscheduled downtime as a result of service fault

Other disruptions may not be added to the availability calculation, such as:

  • Natural disasters

  • Telecom outages

  • Localized outages

  • Malware, cyberattacks, or ransomware: This point is worth expanding, as cyberattacks are becoming more frequent and more severe and can have a major impact on achieving five-nines availability. Some service providers consider these outages "beyond their control" and leave them out of downtime calculations. Good service providers include them. Gold standard services not only offer them but also provide verification, like a SOC2 attestation, which includes availability as a criterion.

  1. Rolling vs. continuous availability:

Some vendors report availability in the previous period (e.g., "past year"); others average previous periods (e.g., "average of past five years"); others measure total availability from service start and synthetically add periods.

  1. How uptime is monitored:

There are many mechanisms to measure availability:

  • Internal tools: log files, system monitoring software/hardware, etc.

  • Third-party monitoring tools

  • Ping testing

  • Customer reports

Not all will report the same numbers. Buyers should ask their sales reps exactly what goes into the calculations for a claimed five-nines uptime.

Verifying five-nines availability claims

Most reputable service providers don't expect customers to take their word on downtime and availability claims. The best providers can produce reports audited by third-party companies, verifying the claims and calculations are true and accurate. Most will at least have unaudited numbers they can show and explain what they include or remove from calculations. It should be a giant red flag if a potential vendor can't or won't back up their claims.

Once you're a customer, you can also use tools like ping testing or monitoring software to track your own uptime. If there's a discrepancy between your availability and claimed availability, don't hesitate to ask your point of contact for an explanation.

How do you get 99.999% availability?

There's no secret to achieving five-nines uptime. It takes:

  • High redundancy

  • Effective failover

  • Fast restart / respawn of services

  • Early failure detection

  • Automation / reduced human error

  • Validated maintenance processes

Some failures will always happen. Getting to five-nines requires anticipating and accounting for those failures, having enough capacity to seamlessly switch over, and building fallback plans into your architecture from the start. Backing into five-nines after the fact is much harder than starting with fine-nines as a goal.

Five-nines is non-negotiable

Or at least it should be. The costs of missing it can be thousands to tens of thousands of dollars per minute of downtime. Equipment and architecture have reached a point where five-nines are both the gold standard for availability and the default assumption. Enterprise buyers should be very wary of any vendor that doesn't offer at least five 9s of availability and should verify the methodology of those that do.

Related articles: