Philip Newton (pne) wrote,
Philip Newton

Shared objects in backups

The other day, I downloaded new versions of my antivirus program.

For safety, I put them into my Dropbox, so that should my computer crash before I got around to installing them, I’d have a copy of the program I paid for.

I was amused to see that backing up the files was nearly instantaneous—presumably what happened was that the Dropbox client took a hash of the file, checked with the server, and saw that someone else had already uploaded those files in the past. So “storing” them didn’t need to involve transferring all the bits over the Internet, but merely creating a connection from my Dropbox file system/object tree to the object that was already present (the moral equivalent of a hardlink).

Same thing when I stored some graphics driver updates just now.

Heh. I find this slightly amusing, though it makes eminent sense for identical files to be stored only once on the backup provider’s end.

(I had seen similar things with Mozy, but it wasn’t as obvious since the backups don’t run in “real time”, but instead a a configurable number of times a day or when started manually, so I usually notice the much faster completion time in such cases. I have seen “already on Mozy servers” in the logs, though—typically for Firefox extensions, given what I back up with Mozy.)

  • Post a new comment


    Anonymous comments are disabled in this journal

    default userpic

    Your reply will be screened

    Your IP address will be recorded