Data splitting

[Brad Fitzpatrick]

Users never see /dev8/0/000/763/0000763408.fid (or
/dev8/0/000/763/0000763408.jpg).

Users would see:

    http://img.domain.pl/my-funny-picture.jpg

And then your app and/or database would map "my-funny-picture.jpg" to the
proper MogileFS key, or perhaps "my-funny-picture.jpg" _is_ the MogileFS
key you used.



On Mon, 6 Aug 2007, drpr0ctologist at gmail.com wrote:

> >
> > Perlbal talks to your webserver, then your webserver talks to your tracker
> > (mogilefsd) or memcached (which has the results of get_paths), and your
> > webserver replies to Perlbal with the locations of the files in MogileFS
> > mogstored node(s), optionally telling Perlbal to cache that mapping
> > internally (Perlbal can internally cache a URL -> @locations for 'n'
> > seconds).
>
>
> Could you give an example of this? I'm a bit confused.
>
> Say user Bob uploads a photo.  The web server receives the file and sends it
> to MogileFS, which then stores it somewhere.  I'm guessing MogileFS returns
> and gives the application some file ID?
>
> Next Bob is forwarded to a page where he now sees his uploaded photo linked
> to a web page.  To get the image URL, the application would need to figure
> out the URL of that file ID.  It tells the web app this URL:
> http://img.domain.pl:7500/dev8/0/000/763/0000763408.fid
>
> The web app then passes that to Perlbal, and Perlbal translates that into:
> http://img1.domain.pl/dev8/0/000/763/0000763408.jpg
>
> Is that how it works?  I'm just guessing here.
>