[Gluster-users] One big vs. many small

[Tobias Wilken]

Hey all,
at our office we've justed discussed the question: Should we use one big
glusterfs share or makes it more sense to use multilpe small shares.

We are going to use glusterfs on multiple virtual instances which hosts
multiple web applications. So we are thinking of providing for every web
application an own glustershare.
Not every web application is on every virtual instance so the glusterfs will
stay small in number of nodes and data to store.

The problem or question are
 - Does the glusterfs overhead eat my bandwith or the number of started
glusterfs processes eat my memory?
 - How is the performance of write duration and read delay compared in the
two situations?
 - Where are the limitations? (Size of glusterfs GB-TB-PB, number of
mounted/started glusterfs's)

I think it depens heavily on the values:
I think with 5 virtual instances, 5 web applications and 100 MB per
application, you'll say: "Boy, take one "big" share and come back, if you
have real problems" :-)
But whats about 500 virtual instaces, 1500 web applications and 500 GB per
application, or the whole thing multiplied by 10 as often you want.

Does someone has experience with similar "problems" or can give hints, why
to use which configuration?!

Best regards
Tobias Wilken