Our initial release of Parallax is now available. Enjoy!
Parallax works by taking a single, cluster-available volume of storage and accessing it though a block management layer. Our block management layer is organized as a radix tree that enables dynamic block remapping. We support read-only sharing of blocks, which are protected with copy-on-write. Through this mechanism, Parallax can perform very low overhead snapshots, and can quickly provision new volumes based on template images.
Parallax is currently in its initial development release, which is based on Xen. For more information, refer to our latest paper, download the current release, or contact us and join the mailing list.
Parallax: Managing Storage for a Million Machines Andrew Warfield, Russ Ross, Keir Fraser, Christian Limpach, Steven Hand. USENIX Hot Topics in Operating Systems (HOTOS), June, 2005. [pdf]
You can also contact us directly by writing to the above address. Public postings are moderated, but the moderator will see all submissions.
Xen sources can be retrieved here. Parallax standardizes against xen-3.2-testing.hg, but we periodically test against unstable releases as well.
Parallax sources can be retreived via mercurial from http://dsg.cs.ubc.ca/parallax/hg/parallax-devel.
Once all sources are in place, follow the README files in the parallax directories.
More information on Mercurial is available here.