Once again, Mother Nature dashes the best laid plans of men. The Friday storm took down Amazon’s cloud data center in Virginia (once again). The storm disrupted power supplies for much of the greater DC area, including Amazon’s US-East-1 region. Apparently the backup power came online too late, and a data center’s worth of virtual servers needed to be restarted. It took Amazon about 24 hours to fully restore cloud services.
This time the cloud crash took down some big names including Netflix, Instagram, and Pinterest. I assume these tech leaders did not choose Amazon’s remote failover capabilities when they set up their cloud based businesses. Simply using the cloud does not insulate anyone from operational problems at data centers. The cloud can be effective provided the service providers and their customers apply normal risk analyses in setting up cloud implementations – remote failover is one response to weather related risks that needs to be considered.
In fairness, it is likely that many on premises data centers were affected by the power outage. These needed to configure their applications for likely risks as well. It’s just that with all the cloud hype, the failures on premises are ignored as are the successes.