AWS S3 SLAs

According to the AWS S3 docs, S3 is designed for 99.99% availability over a given year, but their SLA compensates you when availability drops below 99.9% monthly uptime. That is, you will receive some level of refund (up to 25%) if there is more than 43 minutes of downtime per month.

Specifically, you will get a

• 10% refund if the monthly uptime % is between 99.0% and 99.9%
• 25% refund if the monthly uptime % is less than 99.0%

How is monthly uptime % calculated? Roughly speaking:

Monthly Uptime % = 100% – Error Rate

Where

Error Rate = total # of internal server errors / total # of requests

And is calculated every 5 minutes and averaged over the monthly billing cycle.

Note all this is approximations based on my understanding of their SLA docs. Read for yourself at:

• https://aws.amazon.com/s3/sla/
• https://aws.amazon.com/s3/storage-classes/

It is worth pointing out that all of this is somewhat theoretical though. Infamously on February 28th 2017, AWS S3 was unavailable in the us-east-1 region (Northern Virginia) for 5 hours, obviously not meeting their SLA goals. A good reminder to utilize multi region failover. S3 remained up in the other 13 regions during that time.

See

• Summary of the Amazon S3 Service Disruption in the Northern Virginia (US-EAST-1) Region
• The day Amazon S3 storage stood still

 

Tags: aws, s3, servicelevelagreements, SLAs