[Deploying Sakai] Database vs. Filesystem

[Matthew Jones]

This really isn't a Sakai specific question but more of a general question.
The answers on this stackoverflow question are pretty good and seem
appropriate. [1] Generally for smaller installations it might make sense
because it makes security and maintenance costs easier to store it in the
database. But if you're expecting your system to grow, the database will be
the largest performance bottleneck in Sakai and the hardest to scale. Memory
consumption will go up, as will cpu time by having content in the database.
There also get into possible performance improvments sys admin or dba might
consider such as using SSDs for the database (as it really won't get big if
you don't store content) and then cheap raid arrays for the file storage (as
individual files aren't accessed frequently).

The problem I remember from the list is there is occasionally problems
migrating content out of the database to the file system in different
versions of Sakai 2.x. I think there's conversions for most of the data, but
people have had problems with some of it. So it's really best to have these
questions answered before you go into production like:

How big do you plan on every getting? (If small or same size, database is
probably okay)
Do you plan to ever have a dedicated system administrator/dba? (They will
probably give you better advice)
Do you have local development resources to figure out how to get the data
out if you change your mind later? (Makes it easier to convert if you change
your mind)

I would really advise just starting out file system for any production
instance unless you always plan on only hosting in hundreds of users.

[1]
http://stackoverflow.com/questions/662488/would-you-store-binary-data-in-database-or-in-file-system

On Sun, Jun 20, 2010 at 10:57 PM, Steve Swinsburg <steve.swinsburg at gmail.com
> wrote:

> Hi Drew
>
> Most argue that filesystem storage is better for performance.
>
> However the filesystem structure is still not logical do you can't
> just navigate around the files Sakai creates. The files are just the
> binary bits of the files, the rest of the info about the file (title
> etc) is still in the database. There is also a path field which will
> point to the actual file on disk.
>
> And files created in one site on one day might be in a very different
> location to one created the following day. But Sakai knows how to find
> them.
>
> Try it out on a test instance and you will see the structure.
>
> Cheers
>
> ~steve
> [sent from my abacus]
>
> On 21/06/2010, at 12:19, Drew Blessing <drew.blessing at esu16.org> wrote:
>
> > I am considering storing content on the filesystem instead of in the
> > database.  Can anyone provide some pros and cons with doing so?  I
> > am imagining that management would be easier if the files were
> > logically stored on the filesystem.  I could conceivably navigate to
> > a user's workspace folder and see the files stored there.  These
> > possibilities would be very nice if/when we have a question as to
> > the content a user is storing.
> >
> > Thanks!
> > _______________________________________________
> > production mailing list
> > production at collab.sakaiproject.org
> > http://collab.sakaiproject.org/mailman/listinfo/production
> >
> > TO UNSUBSCRIBE: send email to
> production-unsubscribe at collab.sakaiproject.org
> >  with a subject of "unsubscribe"
> _______________________________________________
> production mailing list
> production at collab.sakaiproject.org
> http://collab.sakaiproject.org/mailman/listinfo/production
>
> TO UNSUBSCRIBE: send email to
> production-unsubscribe at collab.sakaiproject.org with a subject of
> "unsubscribe"
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://collab.sakaiproject.org/pipermail/production/attachments/20100621/df4190ca/attachment.html