We have the pleasure of announcing the first EPEL release of Dynafed, version 1.2.1
The Dynafed system gives seamless access, via HTTP and WebDAV, to a repository that is built on the fly by merging and caching in memory metadata items taken from a number of eventually remote storage systems or endpoints. Dynafed natively supports HTTP, WebDAV, S3 and MS Azure as protocol dialects to communicate with the endpoints. In the S3 and Azure case, the system fully supports managing hierarchical content and applying a uniform authorization scheme to any number of buckets from different providers. The authorization scheme allows not to disclose private S3/Azure keys to the clients. The policies can be based on simple rules or on more sophisticated inline scripting.
From the perspective of a normal user, using HTTP and WebDAV clients, they can access and browse the Dynamic Federation as if it were a large, monolithic filesystem, being redirected to the right host when they ask for a file replica. Dynafed also supports writing. Sysadmins will appreciate the integrated dashboard, providing real time information about the up-ness of endpoints and their history.
The features of Dynafed and its attention to performance and scalability make it an extermely flexible solution to integrate Cloud object storage into an existing computing model, and bridge different authentication domains.
The goal of the system is to provide a multi-site, distributed repository as if it were one entity, without the necessity of maintaining an internal catalogue of all the files it contains. Of course external catalogues of what is supposed to be accessible in a federation will work normally, and such a macro-site can be normally indexed by experiments catalogues.
The packages will become available for public download in the next few days, following the EPEL mirrors propagation.
Special thanks for this important step go to Andrea Manzi and Adrien Devresse.