Ling Liu's SC13 paper "Large Graph Processing Without the Overhead" featured by HPCwire.
Another list highlighting Open Source Software Releases.
Second GraphLab workshop should be even bigger than the first! GraphLab is a new programming framework for graph-style data analytics.
A case for scaling HPC metadata performance
Proc. of the Seventh Parallel Data Storage Workshop (PDSW12), co-located with the Int. Conference for High Performance Computing, Networking, Storage and Analysis (SC12), November 2012.
Swapnil Patil, Kai Ren and Garth Gibson
Carnegie Mellon Univeristy
We envision a scalable metadata service with two goals. The first goal – evolution, not revolution – emphasizes the need for a solution that adds new support to existing cluster file systems that lack a scalable metadata path. Although newer cluster file systems, including Google's Colossus file system , OrangeFS , UCSC's Ceph  and Copernicus , promise a distributed metadata service, it is undesirable to replace existing cluster file systems running in large production environments just because their metadata path does not provide the desired scalability or the desired functionality. Several large cluster file system installations, such as Panasas PanFS running at LANL  and PVFS running on Argonne BG/P , , can benefit from a solution that provides, for instance, distributed directory support that does not require any modifications to the running cluster file system. The second goal – generality and de-specialization – promises a fully, distributed and scalable metadata service that performs well for ingest, lookups, and scans. In particular, all metadata, including directory entries, i-nodes and block management, should be stored in one structure; this is different from today's file systems that use specialized on-disk structures for each type of metadata.
FULL PAPER: pdf