[BlueOnyx:10654] Re: Ploop

[Michael Stauber]

Hi Steffan,

> Is this something for Aventuring
> http://wiki.openvz.org/Ploop
> I read a lot of positive results

Yeah, I follow the discussions there as well to stay in the loop about 
interesting ideas and approaches.

I looked at Ploop and experimented a little with it on 6106R.

Naturally many users on the OpenVZ mailing list are comfortable using bleeding 
edge code. And there is quite some infatuation with experimental file systems.

Last year GlusterFS (http://www.gluster.org) was "the next big thing" for 
OpenVZ, this year it is Ploop. I also checked GlusterFS, as it was a 
particularly nice alternative to DRBD for clustered systems. 

But at this time there is one thing that both Ploop and GlusterFS have in 
common and it really affects the ability to market it as a fundamental 
component in a commercial environment:

Both are experimental and not yet finished.

Don't get me wrong: Ploop is indeed interesting in so far as it allows to 
compartmentialize the VPS's in a fashion that isolates them better from the 
master node. That has a few advantages over basing a VPS directly on a 
"regular" filesystem - no doubt. But all in all the advantages are (in my 
opinion) not substantial enough to outweight the risks of an early adoption of 
Ploop.

When the very same filesystems developers that created it say "It's not yet 
ready", then I'm of course really reluctant to offer it to paying customers. 
Especially if it is one of the most (if not THE most) crucial component: The 
very own filesystem that stores the invalueable data that you plan to host.

Hey, I like bleeding edge code. But when it comes to filesystems, then I 
really, really follow a no nonsense and very conservative guideline. I still 
use EXT3 even long after EXT4 has been deemed "stable" and ready for 
production. 

Both GlusterFS or Ploop offer some nice new features that I'd love to have for 
this or that reason, but in my view the unclear desaster recovery process and 
the warnings straight form the developers that "the code is not yet fully 
ready" let me stay clear of it outside of some controlled testing environments 
that only have unimportant data on them.

However: I wouldn't feel comfortable hosting important business data on a 
Ploop filesystem or a GlusterFS at this time. Let us give it some more time to 
mature ...

-- 
With best regards

Michael Stauber