Initially I was thinking the disaster recovery scenario. For that everything just in Glacier would be acceptable. If we want to start serving files from Amazon too (via http redirector etc), we would need some way to know when a file is in Glacier and when it is in S3. So when a file requested is only in Glacier, we serve the file from a different replica and also send a 'restore' request to Glacier so that the file can be served from S3 next time it is requested. This could mean a lot of back and forth but we can adjust the migration to more user requests per 'restore' request. So S3 would be a cache of frequently/recently requested files. -- bye, pabs http://wiki.debian.org/PaulWise
Attachment:
signature.asc
Description: This is a digitally signed message part