Amazon S3 offers two plans:
Storage
(Designed for 99.999999999% Durability)
and
Reduced Redundancy Storage
(Designed for 99.99% Durability)
Designed to provide 99.999999999% durability and 99.99% availability of objects over a given year.
So if I have 10 000 files I can expect to loose one in average with the second plan per year? am I interpreting that correctly?
Edit: Maby I have to put it more clearly:
How do you interpret especially the “designed for” part. If for example offer 99,9% availability I guarantee that and pay a penal for every hour more or something like that.
But if I design the system for 99,9% availability, I choose the parts with the knowledge that the system will probably have a downtime of 0.1% on statistical average.
That does not necessarily mean that I guarantee anything. Its just what the system is designed for…
RRS will survive the total loss of one data centre, as well using multiple drives, etc. Regular S3 storage will survive the loss of two.
The 99.999999999% durability figure is a bit pointless – they’re quoting something less than the the probability of a K/T event sized comet hitting Earth, and if that happened we wouldn’t be worrying about our S3 data.
The probability of a political or economic event taking down Amazon is a much higher than the risk of three AWS regions being simultaneously destroyed, so I reject that notion that data in S3 “doesn’t need to be backed up” although it is good to know that the system has been designed to survive most eventualities.
The risk of having a system you have connected to your AWS account compromised, and losing data that way, is higher again, so always keep offline backups.