Payments are experiencing issues due to temporary restrictions in Russia. If your payment does not go through, please submit a support request.Our support team is available 24/7 — we are always here to help with hosting and server issues.We are now accepting requests for dedicated server rental and colocation services in our data center.Reminder: we recommend enabling backups for additional data protection.A new VPS/VDS lineup with NVMe storage and improved performance is now available.Maintenance work on some servers has been completed. All services are operating normally.
Article4 min readViews1

99.9% Availability Equals 43 Minutes of Downtime in 30 Days

Converting percentages to minutes separates hosting metrics from store performance. A sample calculation helps set a meaningful availability target and clarifies exactly what needs measuring.

Comments 0

Network switch with connected cables and a spare cable nearby. Generated illustration.
In this article

In 30 days, there are 43,200 minutes. If availability is measured by time and the target is set at 99.9%, the remaining 0.1% of the period equals 43 minutes and 12 seconds of downtime. This figure may seem small until those minutes coincide with the start of a sale.

This is an arithmetic example, not a promise from a specific provider or a forecast of a future outage. A percentage defines the proportion of downtime within a chosen period. It does not indicate whether the downtime will be one long outage or many short ones, or whether it will occur at night or during peak demand.

An Additional Nine Changes the Scale of the Task

The formula for the time-based metric is simple: multiply the period duration by the proportion of unavailability. For 30 days, a target of 99.95% leaves 21 minutes and 36 seconds, while 99.99% leaves 4 minutes and 19.2 seconds. Moving from 99.9% to 99.99% reduces the allowable downtime by a factor of ten, even though visually only one additional nine is added to the number.

Training downtime budget for 30 days: 99.9% — 43 minutes 12 seconds; 99.95% — 21 minutes 36 seconds; 99.99% — 4 minutes 19.2 seconds.
Sample calculation for the time-based metric over 30 days. Specific SLA conditions may differ.

Imagine a hypothetical store with a target of 99.9%. A single stoppage lasting 30 minutes would already consume approximately 69.4% of its monthly downtime budget. Only 13 minutes and 12 seconds would remain before reaching the limit. At a target of 99.99%, that same stoppage would exceed the budget by nearly seven times. No additional hours of successful operation within the same thirty-day period can turn such a result into compliance with a stricter target.

A 31-day month has a different denominator: 44,640 minutes. For 99.9%, that equals 44 minutes and 38.4 seconds. Therefore, a report must not take a percentage from a calendar SLA and minutes from a hypothetical month without explaining the difference. Rolling 30-day periods, calendar months, and working hours yield different calculations.

First, determine what counts as unavailable.

An SLA defines the service provider's obligations. An internal team goal, an SLO, may describe a different object and be stricter. For example, the agreement might cover a virtual machine, while the store is concerned with the ability to complete a checkout. You cannot automatically apply the availability percentage of one service to the entire customer journey.

In practice, it is useful to list four items: the measurement object, the failure criterion, the period, and exceptions. These may be minutes of resource unavailability or the share of successful requests. Scheduled maintenance, minimum incident duration, and architecture requirements are handled as defined in the specific agreement. There is no single set of exceptions applicable to all hosting providers.

Yandex Cloud, for example, publishes general SLA terms and separate service levels for services. The very fact of such a division is a reason to read the terms of the selected product rather than transferring a figure from a neighboring service. A percentage indicated in advertising without a defined scope does not provide a comparable selection criterion.

If a metric is calculated per request, it cannot be directly converted into minutes. With 100,000 considered requests, a 99.9% target leaves 100 failed requests. It is unknown when they occurred or how long the degradation lasted based solely on this sum. A night minute with ten requests and a day minute with a thousand requests have the same weight in a time-based metric but different weights in a request-based metric.

What to buy for four minutes

A stricter goal requires changing recovery methods, not just the tariff. If failure detection takes five minutes, then a person spends ten minutes searching for access and another twenty restoring the system, the monthly budget for 99.99% availability can be exhausted before repairs even begin. Here, it is useful first to reduce specific delays, and only then discuss redundancy.

A second server helps only with a clear failover and available data. Two applications depending on the same database retain a single point of failure. A backup solves the data recovery problem, but the time required to deploy it does not disappear. A strict goal must rely on a tested scenario, including failover errors and the return to normal mode.

At the same time, the same architecture is not economically justified for every small store. You can set separate goals for catalog browsing and checkout, define critical hours, and establish recovery priorities. This is not permission to hide inconvenient failures from reports: measurement rules must be defined in advance and applied consistently.

The practical outcome of the calculation is a concrete question for the team: what is the total allowable downtime for the selected operation, and can we meet this target during a typical incident? If the answer is not yet known, the next useful step is to measure detection and recovery on a safe training scenario. Purchasing an additional nine in the service description will not close this gap.

Discussion 0

Share your experience and ask questions. Comments without links appear after editorial review.

No comments yet. Start the discussion.