We are using Amanda to send backups to S3. Each host is acting as its own Amanda "server".

All hosts are currently using the same S3-bucket -- each has its own subfolder under it. The dumps are saved with "REDUCED_REDUNDANCY" -- it is 20% cheaper and we consider it unlikely, that both the original host and the backups will disappear at once.

We do not currently specify, where the bucket is stored -- using whatever Amazon does by default.

Our EC2 systems are spread across different regions, however, and we are wondering, if, perhaps, we should use multiple buckets -- one explicitly created in each region storing dumps of hosts from another region. For example, make us-west-1 hosts send their dumps to the us-west-2 bucket (or even to us-east-1) -- to keep the backups available should Amazon lose a particular region entirely (both the EC2 virtual machines and S3 data).

Would that be useful, or is S3-data already mirrored to multiple regions by default? Will there be increased billing costs -- it is our understanding, that writing to S3 is free, and storage costs the same -- is that correct? Latency is of no particular concern -- as long as bandwidth remains decent.


If you're using the default region US Standard, your files get (eventually) stored in both us-west-2b and us-east-1. This is the only region with this redundancy. All other regions store only to servers in that region, so you would have some additional safety by using a different region for backups - if a meteor takes out your Oregon servers, having backups in Virginia will be a happy thing to have planned ahead with.

Your bandwidth costs will change doing this, though, as traffic to S3 from a different region is billed at the "EC2->Internet" rates. Using US Standard will give you the best of both worlds - free transfer to S3 and the geographic redundancy.


  • 2
    Are you sure about US Standard being replicated across the continent? I've heard directly from AWS Support that, in no uncertain terms, this is not the case. The object can be accessed from both locations without bandwidth penalty, and the "non-primary" location will cache the object for some time, but not permanently. – EEAA Jul 12 '14 at 3:50
  • 2
    This is not accurate. Us-west-2 is it's own region and not part of us standard. docs.aws.amazon.com/AmazonS3/latest/dev/LocationSelection.html. Us-west-2 is Oregon and us standard is Virginia and Pacific Northwest. – imperalix Jul 12 '14 at 5:39
  • @EEAA I've linked to the relevant docs. – ceejayoz Jul 12 '14 at 14:50
  • @imperalix I'd figured as Oregon is in the PNW that it's there. A look at the AWS infrastructure page indicates they also have a Seattle location. Not clear which they use from the docs I've seen. – ceejayoz Jul 12 '14 at 14:52

s3 buckets are normally set for a particular region. When creating a bucket you can set it's region. You can set cross region replication as well. I don't think cross region replication is set by default.

With s3 you get charged for the amount you store, and the data transferred. You best look at the actual rates here: https://aws.amazon.com/s3/pricing/. There billing for s3 is generally cheap, but complicated.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.