An Innovative Tool for IO Workload Management on Supercomputers

Presenter(s): Si Liu (TACC)

Presentation Slides

Modern supercomputer applications have been driving a high demand for capable storage resources in addition to fast computing resources. However, these storage systems, especially parallel shared filesystems, have become the Achilles' heel of powerful supercomputers. Single user's improper IO work can easily result in global filesystem performance degradation and even unresponsiveness. In this project, we developed an innovative IO workload managing system that optimally controls the IO workload from the users' side. This system will automatically detect and restrict improper IO workload from supercomputer users to protect parallel shared filesystems.

The Brain Image Library

Presenter(s): Derek Simmel (PSC)

Presentation Slides

The Brain Image Library (BIL) is a national public resource enabling researchers to deposit, analyze, mine, share and interact with large brain image datasets. As part of a comprehensive U.S. NIH BRAIN cyberinfrastructure initiative, BIL encompasses the deposition of datasets, the integration of datasets into a searchable web-accessible system, the redistribution of datasets, and a High Performance Computing enclave to allow researchers to process datasets in-place and share restricted and pre-release datasets. BIL serves a geographically distributed user base including large confocal imaging centers that are generating petabytes of confocal imaging datasets per year. For these users, the library serves as an archive facility for whole brain volumetric datasets from mammals, and a facility to provide researchers with a practical way to analyze, mine, share or interact with large image datasets. The Brain Image Library is a operated as a partnership between the Biomedical Applications Group at the Pittsburgh Supercomputing Center, the Center for Biological Imaging at the University of Pittsburgh and the Molecular Biosensor and Imaging Center at Carnegie Mellon University.
In this talk, I will briefly review the characteristics of the data that the Brain Image Library will store, and the infrastructure we are building at PSC to ingest and manage the data for access.