SLASH2 at SuperComputing 2013

SLASH2 was featured in the PSC booth at SC13 in Denver, CO. The demo included a graphical interface for passersby to actually drive a SLASH2 deployment file system and manage replicas and watch data coherency in realtime, all the way remotely back to Pittsburgh!

Handouts covering high level system internals as well as handouts for the Data SuperCell and Data ExaCell were also being distributed. The material seemed to be well received by many!

Replication starvation issue resolved

A first run in the guts of the SLASH2 metadata server replication engine left an issue with starvation of replication workloads queued up by users. A classical starvation issue, a single user can inject many work items and deny service to other users as pieces of work are selected at random, instead of randomly throughout the user class.

Student interns Chris Ganas and Tim Becker provided some exploration and improvements in the code to do fair sharing in the engine efficiently and in a manner that can scale to large system workloads. Good work and thanks guys!

Replication priorities

Support for user- and system-level prioritization is almost complete. The lower level engine has had support for this feature for some time but some of the higher level guts had not been finished. There are still a few challenges left but we are getting very close now…

Replication policy inheritance

The HEAD branch of SLASH2 has been recently updated with support for user replication policy inheritance. What this means is that via msctl(8), a user can set preferred residency settings on a directory and all files (and subdirectories) created thereunder automatically inherit this policy. This means no further commands are necessary to manage residencies after creation. And full support for persistent replications is available via this mechanism as well. Happy (easy) data managing!

PSC awarded NSF grant for new storage system

The National Science Foundation (NSF) has approved a grant to Pittsburgh Supercomputing Center (PSC) to develop a prototype Data Exacell (DXC), a next-generation system for storing, handling and analyzing vast amounts of data. The $7.6-million, four-year grant will allow PSC to architect, build, test and refine DXC in collaboration with selected scientific research projects that face unique challenges in working with and analyzing “Big Data.”

The full press release is available here:

http://www.psc.edu/index.php/newscenter/2013/872-psc-lands-76-million-data-exacell-grant

SLASH2 will be just one of the software interfaces available to access the capabilities of the new storage resource. The system will focus on much higher performance demands than previous deployments featuring SLASH2, so it will be excited to see how fast we can go and how far we can scale out!