[Labs-l] beta cluster down (cause of GlusterFS)

[Ryan Lane]

On Mon, Apr 22, 2013 at 12:06 AM, Petr Bena <benapetr at gmail.com> wrote:

> <futurama>HahahawmHAhahahmuahah hahaa haaaa heeee...</futurama>
>
>
> who it was telling me that nfs suck and I am really n00b if I am using
> instead of brilliant gluster? :P
>
>
NFS *does* suck. It's an acceptable solution short term, but we're going to
eventually hit scaling issues with it and will have to replace it in the
future. There's a reason everyone in the world is looking at and working on
scale out solutions right now.

Having an NFS solution inside of a project is not a scalable solution for
multi-tenancy, as it would require an instance for every project. It also
adds ordering to how instances are started and stopped and adds a single
point of failure to every project.

Overall using an NFS server is not a great solution, but we don't really
have an alternative right now. If we didn't care about ease of use, we'd do
what most cloud providers do by just providing block storage and letting
everyone handle things themselves, since that's "the cloud way".

- Ryan
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.wikimedia.org/pipermail/labs-l/attachments/20130422/c0248235/attachment.html>