[Openstack-operators] Management of NAS (NFS/CIFS shares) in OpenStack

Blair Bethwaite blair.bethwaite at gmail.com
Thu Nov 22 00:29:38 UTC 2012


Hi Ben,

> Date: Wed, 21 Nov 2012 15:52:46 +0000
> From: "Swartzlander, Ben" <Ben.Swartzlander at netapp.com>
<SNIP>
> NetApp is interested in providing a common way to provision and attach
> NFS and CIFS shares to VMs in an OpenStack environment. We presented our
> ideas at the OpenStack summit last month and we have submitted a patch
> to Cinder that demonstrates one way to implement these features.

I was at your session and listened with keen interest. This is a great
feature to introduce and something that I'm sure will be welcomed by the
operator and user communities! Efficient, automated access to existing
large (file-based) data-sets and high-performance shares that support
data-intensive workflows are key drivers (actually blockers at the moment)
for uptake in the research sector (at Monash University we want to get at
an existing GPFS). Currently, a lot of users and devs working with/on
cloud-services see volumes (block-storage) as their only option for "large"
data support, when what they really want is a file-system.

It's unfortunate that a good deal of that session was dominated by
discussion about whether this ultimately belongs in Cinder or not, rather
than drilling down into some of the more interesting possibilities - you
may recall I mentioned parallels with a session from USC-ISI regarding
guest access to high-performance file systems.

> We
> believe that extensions to the Cinder API are the most logical and least
> disruptive way to bring these features to OpenStack, but discussions at
> the conference show there are differences of opinion on this topic.

I can see this from both points-of-view. But I have to agree with you at
present, I think pushing this into a separate service from inception would
make for undue overhead without an existing community of users and devs to
help (which will be found more easily if the feature is already in Cinder),
not to mention introducing a (relatively) long and involved delay in
getting such a new service into core. And I think Cinder should be allowed
to evolve without needing to be singularly focused on block-storage only,
rather it should be about exposing _storage_ as a service for guests. I
would think a number of storage vendors (beyond Netapp) would share this
view too, preferring to work with one project to add support for their
various block&file-based products. If it turns out FSaaS outgrows its'
boots and becomes unmanageable alongside BSaaS then they can be split
(that precedent has been set) - that would actually be a healthy problem to
have IMO.

-- 
Cheers,
~Blairo
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-operators/attachments/20121122/026a3493/attachment.html>


More information about the OpenStack-operators mailing list