Open Stack

> just discussion regarding ceph as platform for swift and cinder, and
> also a replacement.
>
> because we can run swift and cinder also without ceph
> and why should ceph as both platform,

As in the true linux spirit you have options.
Make sure you know what your requirements are and see what fits best for your use-case.

There are some good reasons for using Ceph for both Swift and as a Cinder backend (you still make use of the Cinder APIs)
* Having one large data pool makes sure you use space efficiently.
* Fewer technologies to get familiar with.

RadosGW vs Swift:
* You can no longer make use of Swift pipeline / middleware
 This is very flexible and can add some nice features
* Not so sure about scalability with a lot of (small) files in with Ceph
Swift has a somewhat similar issue but it seems to be harder to improve on Ceph.
With Swift you replace the container servers with SSDs, I do not *think* you can do something similar with the metadata databases in Ceph.
It looks like Ceph is working on improving this ( http://wiki.ceph.com/Planning/Blueprints/Submissions/rgw%3A_bucket_index_scalability )
* Swift seems more mature (debatable)
Just from following the maillinglists I get this impression...

Using Cinder with Ceph vs Cinder with other systems:
* You will always need Cinder, it provides the required API
* Ceph is awesome as a backing store with very good performance in distributed loads. 
E.g. lots of VMs doing some IO.
However you might not get the desired performance from a single VM doing lots of random write IO.
(especially if you do not allow write caching which makes sense for e.g. databases)
* Ceph does not have all "enterpricey" SAN features (yet)
>From a architecture point of view Ceph is just awesome, however there are a lot of challenges left.
Just to name a few features that are not there yet: data de-duplication, throttling, asynchronous multi-dc replication (for rbd)

The last thing to take in account is that Ceph is still evolving and I think in the long term Ceph proposition will be very hard to beat.
It might not fit your exact needs today but in a few years there will be very few reasons to buy a SAN...

Cheers,
Robert van Leeuwen