A collection of large cloud outages collected by year which to serve as a data basis for the argument against believing in cloud SLAs. Not that I'm against using the cloud, but knowing there will be several days outage a year with a certain probability despite any promised SLA is important.

To understand which outages are listed and why others are not. The criteria are roughly

• outage is global
• or very long >1h
• or causing a complete service loss in a region