Can we stop with the uptime percentages?
summary
Uptime percentages are an inefficient way to communicate service reliability to the general public because they fail to convey the magnitude of downtime. While infrastructure experts understand the significance of 'nines,' the average user perceives a small percentage difference as negligible. Shifting to a time-based measurement, such as total hours of downtime per month, provides a clear and intuitive understanding of a service's stability. This change would make status pages more transparent and honest about the actual impact on users.
key points
- Percentage-based metrics are deceptive because the difference between 99.9% and 99.99% represents a tenfold increase in reliability, which is not intuitive to non-experts. This creates a gap in understanding between the infrastructure engineers and the general user base.
- Status pages now serve as the broader public interface for a service, meaning they should prioritize human-readable information over industry-standard compliance metrics. Measuring downtime in hours or minutes makes the reliability of a service immediately clear to anyone without requiring mental math.
- Combining both the time-based impact and the percentage for transparency ensures that technical users still have the same data while removing the ambiguity of high percentages. This approach translates an abstract number into a real-world experience of service interruption.
who should read it
Product managers, developers, and site reliability engineers who design public-facing status pages and communicate service health to their users.
discussion skeptical
consensus
Commenters largely agree that percentage-based uptime metrics are misleading and downplay the actual impact of outages. They argue that expressing downtime in hours makes the loss of productivity and revenue more tangible, especially for critical infrastructure like GitHub. There is a strong sentiment that official status pages are often untrustworthy or 'cooked,' leading users to rely on third-party tools like Downdetector.
top insight
The concept of 'uptime' is fundamentally flawed for distributed systems because a single number cannot account for partial failures, varying user impact, or the timing of outages. A 98% uptime figure looks acceptable on paper, but if the downtime occurs during peak business hours, the effective disruption to the user is significantly higher than the average suggests.
Generated by an LLM from the article and thread. Check the sources before relying on it.