The Cost of 99.9% Uptime in AWS
The Amazon outage has renewed the conversation around the value of abstraction vs. the sunk cost of 99.99% infrastructure
One of the significant conversations coming out of the Amazon outage is the effort needed to achieve 99.9% uptime in the public cloud. Amazon is transparent that a single region provides only 99.9% uptime. I got into a spirited conversation on LinkedIn about how the outage isn’t a big deal. If you’ve built a service based upon a single AWS region, you’ve accepted the resulting uptime.
However, this may be an oversimplification. It’s tough, if not impossible, to build the types of services AWS provides with private infrastructure. The sheer velocity that the public cloud offers has been a game-changer.
However, enterprises have spent years and millions of dollars building highly redundant infrastructure designed to make monolithic applications resilient. That investment becomes a hidden benefit to every additional application you add to your private infrastructure.
Therefore, pay as you go public cloud means you have to pay for 99.99% uptime for each application you build. It has put many customers in the uncomfortable position of reassessment of risk. Does a customer embrace agility and lose resiliency to keep costs in line? Or does a customer embrace the agility and pay on an app-by-app basis for resiliency?