Some “powerful” results came out recently from Uptime Institute‘s Uptime Intelligence 2024 Annual Data Center Resiliency Survey.
On the Bright Side…
Concern has been growing across the Data Center industry that the overwhelming power demands caused by overall increased computing across industries and particular to gen AI will lead to more frequent – and more serious – outages. Surprisingly, Uptime Institute‘s 2024 Annual Data Center Resiliency Survey found that both the frequency and severity of data center outages in fact DECREASED in 2023 from prior years. The charts below show a clear downward trend over the past three years.
A 14 percentage point decline in reported outages between 2021 and 2023…
And a 10 percentage point decline in outages characterized as either serious or severe.
The survey determined that outages are decreasing relative to the overall rise in IT because public and private organizations have been vigilant in their spending, training and compliance practices to address the many growth challenges created by the proliferation of new technology. That’s the good news.
Power Struggle
The bad news is that Uptime’s annual surveys show that things are only getting harder when it comes to the most common factor behind significant outages: disruptions to on-site power distributions. In fact, the study revealed that IT hardware can’t handle voltage fluctuations or power losses that last more than even fractions of a second! The below exhibit summarizes the top causes of significant outages.
Just over half of survey respondents cited power as the cause of their most recent serious outages, with cooling in distant second place at 19%. What’s more, as third-party providers have played a larger role in the DC ecosystem, so has their responsibility in disruptions, rising from 5% in 2020 of total outages cited to 9% in 2023.
The Human Factor
Let’s not forget that machines, no matter how sophisticated, are still operated by people (for now, anyway!). Sometimes these people are fatigued, distracted, error-prone, insufficiently trained or just don’t show up. Uptime used 25 years of data to analyze how many times human error contributed directly or indirectly to all downtime incidents and found the number ranged from 66-80%!
Specific to the 2023 survey, the exhibit below summarizes the survey response to the most common causes of major human error-related outages.
The takeaway:
Solving the power issue starts with outages, but the implications extend well beyond data center resiliency. Consistent and reliable access to power is at the heart of the sustainability, development and growth of the IT infrastructure that supports data centers around the world and the growth and prosperity of so many industries and societies around the world. According to the survey, “resiliency consistently ranks among the top concerns among industry stakeholders and is a strong driver of investment.” Let’s hope this shared vigilance – and investment – can outrun the demand.
