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.
Distributed Cloud Storage Services with FleCS Containers
5th OpenCirrus Summit, Moscow, Russia, June 2011.
Hobin Yoon, Madhumitha Ravichandran, Ada Gavrilovska, Karsten Schwan
Georgia Institute of Technology
There are limits to the ability to migrate or deploy applications across geographically distributed/loosely coupled cloud resources, requiring substantial data movement and/or uniformly visible and accessible storage services across such distributed infrastructure. To address these issues, we propose and explore the utility of FleCS – an approach for providing FLExible Cloud Storage services in distributed systems. FleCS provides storage containers as a cloud-level abstraction that uniquely identifies a subset of storage resources and their associated attributes. Attributes determine certain container properties, including those concerning data replication and consistency, thereby creating opportunities to pay those costs only for the state/data which require them. FleCS exports to cloud applications an object-based storage API that allows them to request the 'right' types of storage, and to correspondingly group/classify their data. Sample uses go beyond the established notions of application-provided or derived hints to classify the 'hotness/ coldness' of data and/or to provide better energyefficient storage services, to also include applicationspecific notions of data consistency and update strategies.
FleCS and several types of storage containers are realized for a prototype platform consisting of groups of nodes, virtualized with the Xen hypervisor, with distinct storage targets, each managed by a separate NFS server. Evaluations use benchmarks based on popular cloud applications. A future target platform for evaluation is a distributed OpenCirrus cloud infrastructure spanning multiple data centers.
FULL PAPER: pdf