SwiftStack's Joe Arnold and John Dickinson chose the Austin Summit and a low-key, #vBrownBag venue, to come out of closet with PROXY-FS (also spelled as ProxyFS), a tightly integrated addition to OpenStack Swift, which provides a POSIX-ish filesystem access to a Swift cluster.
Proxy-FS is basically a peer to a less known feature of Ceph Rados Gateway that permits accessing it over NFS. Both of them are fundamentally different from e.g. Swift-on-file in that the data is kept in Swift or Ceph, instead of a general filesystem.
The object layout is natural in that it takes advantage of SLO by creating a log-structured, manifested object. This way the in-place updates are handled, including appends. Yes, you can create a manifest with a billion of 1-byte objects just by invoking write(2). So, don't do that.
In response to my question, Joe promised to open the source, although we don't know when.
Another question dealt with the performance expectations. The small I/O performance of Proxy-FS is not going to be great in comparison to a traditional NFS filer. One of its key features is relative transparency: there is no cache involved and every application request goes straight to Swift. This helps to adhere to the principle of the least surprise, as well as achieve scalability for which Swift is famous. There is no need for any upcalls/cross-calls from the Swift Proxy into Proxy-FS that invalidate the cache, because there's no cache. But it has to be understood that Proxy-FS, as well as NFS mode in RGW, are not intended to compete with Netapp.
Not directly, anyway. But what they could do is to disrupt, in Christensen's sense. His disruption examples were defined as technologies that are markedly inferior to incumbents, as well as dramatically cheaper. Swift and Ceph are both: the filesystem performance sucks balls and the price per terabyte is 1/10th of NetApp (this statement was not evaluated by the Food and Drug Administration). If new applications come about that make use of these properties... You know the script.