Amazon Web Services was the most reliable of the major cloud infrastructure operators in 2014, according to new year-end data from provider comparison service CloudHarmony, but it trailed many small regional and specialty rivals in base uptime, and lagged Google in the storage area.

Amazon boasted a comfortable lead over its largest competitors in the compute category with total downtime of less than three hours over the past 12 months. Its two largest rivals – Google and Microsoft – didn’t perform nearly as well despite investing aggressively in their platforms, according to statistics from CloudHarmony’s CloudSquare monitoring service. Google trailed more than half a dozen other providers – including much smaller fish – with 4.43 hours of downtime across some 71 outages, while Microsoft’s service was inaccessible nearly ten times as long as that in the past year, putting it close to the bottom of the list. Even so, Microsoft provided 99.9336 percent availability.

Amazon’s dominance was not complete. It trailed Google in the storage category with 22 outages lasting a combined two-and-a-half hours. Microsoft also had problems in storage as the result of high-profile outages that its dragged down its overall position in the compute department.

Uptime is an important metric for the maturity of a cloud platform, but despite a multi-year head start and billions of dollars of investment, the top three operators don’t have as big of an edge as you might think. On reason is because the services they provide are becoming more specialized and complex, and thus more fragile.

As Amazon cloud mastermind James Hamilton detailed in a recent interview on theCUBE, the company runs each one of its services on a specially designed hardware stack optimized for that particular workload down to the component level. Google and Microsoft have implemented the same approach in their own data centers.

This approach is more reliable than the often-Byzantine architectures that exist within many corporate data centers, but it also creates complexity as additional services are added. Smaller providers such as the recently-funded DigitalOcean Inc., which beat Azure in virtual machine availability, say they’re reliable because they rely entirely on off-the-shelf hardware.

DigitalOcean has emerged as the third largest cloud provider thanks to a narrow focus on making it as simple as possible for developers to spin up hardware resources over the web. Other niche players have a similar opportunity to leverage the top three cloud operators’ size against them, and many are already doing so, according to the CloudSquare data.

That opening won’t last forever, though. As Hamilton explained to SiliconANGLE, Amazon experienced its fair share of growth pains in previous years but gradually improved its operations over time, drawing lessons from every new service and feature. Google and Microsoft are moving through the same learning cycle, slowly closing the gap on availability.

Photo via Pixabay