Open Stack

On Wed, 2015-05-27 at 10:28 +0200, Vincenzo Pii wrote:
> Let's assume that a software-defined storage system provides extremely
> reliable filesystems to be used as devices in Swift rings.
> 
> 
> For the sake of reliability, this underlying storage can be considered
> to provide data integrity and availability in 100% of the cases
> (strong assumption, but helpful to get to the point).
> 
> 
> Performance in this context is really not a concern.
> 
> 
> To avoid reducing the storage efficiency, an idea could be to use
> swift with number of replicas set to 1 in this context.
> 
> 
> My question is the following: when performance doesn't matter and
> reliability is taken care of "below" swift (so swift will always
> manage to read/write an object as devices will always be consistent
> and available), are there other aspects that should be considered if
> swift runs with a replica counter of 1?
Hi Vincenzo,
The Swift-on-File [1] project works as you described, the difference is
that we implemented our own object server to store the data on-disk a
little bit differently, but you could still use the default swift object
server.

One important aspect to consider is where your account and container
databases are going to be stored. You probably don't want to have just
one replica of your account/container databases and have that accessed
by multiple services, so you might want to consider deploying them the
default way (writing to JBOD disks and having multiple replicas)

[1] - https://github.com/stackforge/swiftonfile

> 
> 
> Many thanks,
> Vincenzo.
> _______________________________________________
> Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
> Post to     : openstack at lists.openstack.org
> Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack