Amazon S3 as remote/backup/shared filestore

[ofrxnz]

Today i had the idea that their may be a use for being able to use Amazon's S3 system as a the backup filestore or the shared filestore for a cluster. 

I know there may be some bandwidth issues and i probably couldn't use it due to federal regulations but for a small shop it would handle all of their file backups. 

Clearly the primary store would have to be on the local drive but a remote store like this may prove useful to someone.

Anyone have any input on this?

Adam

P.S. Amazon also has the DB and system service but i think there would be too much of a performance issue unless  they were used as a backup or the shop had no LAN business only internet based Alfresco.

[pmonks]

Luis (http://blogs.alfresco.com/wp/luissala/) implemented a prototype S3-backed content store (http://dev.alfresco.com/resource/docs/java/repository/org/alfresco/repo/content/ContentStore.html) some time ago - it works much the same as the replicating content store (http://dev.alfresco.com/resource/docs/java/repository/org/alfresco/repo/content/replication/Replicat...) but replicates to S3 rather than a vanilla filesystem.

I don't know if he's put the code out on the forge or whatever, but I'll ping him and find out what his thoughts are on making it available.

Also, this only replicates the files themselves to S3 - it doesn't take the Alfresco database into account (which would also be required if this was to be a true backup solution - see http://wiki.alfresco.com/wiki/Backup_and_Restore for all of the requirements for successfully backing up an Alfresco repository).

Cheers,
Peter

[ofrxnz]

Peter and Luis

Thanks for the quick response and great tips!!  I will probably dig further into this. 

I knew it wouldn't handle the database, but since MySQL's INNODB supports hot backup, it would not be to difficult to script a warm/hot backup and push it to S3.  then it would be a matter of snapshoting or copying the filestore at roughly the same time so the DB and File system represent each other.

I wonder if S3 could snapshot  a file system? 

We have geographically separate sites and were looking at placing an alfresco server at each location.  I thought S3 would be a possibility to mitigate any issues bandwidth or connections at a site.  If i can get over any regulatory issues. 

Anyways, it is an interesting concept and I believe i have to play around.

Adam