Evaluating ScaleMP virtual shared memory architecture for use in HPC
Project Information
- Discipline
- Computer Science (401)
- Subdiscipline
- 11.04 Information Sciences and Systems
- Orientation
- Research
This work will provide an unbiased evaluation of the ScaleMP's vSMP foundation for use in supporting mid-tier scientific computing. We expect this technology to be especially useful to users who need to scale their application quickly and are unable to completely re-write their code. Furthermore, this work could significantly enhance many big-data applications by allowing scientists to move databases on traditional tertiary storage to 5.6TB of system memory, which could effectively speed up computation by many orders of magnitude.
Broader ImpactsThis evaluation of the ScaleMP vSMP foundation could advise future large-scale cyberinfrastructure that aim to support shared memory systems. Furthemore, this work could help guide resarchers on how to best use the new shared memory architecture with minimal effort.
Project Contact
- Project Lead
- Andrew Younge (ajyounge)
- Project Manager
- Andrew Younge (ajyounge)
- Project Members
- Scott McCaulay, Chaitanya Gogineni
Resource Requirements
- Hardware System
-
- Not sure
This will use the ECHO system with ScaleMP's vSMP foundation.
Scale of UseWe will scale up to all 16 ndoes of Echo, and may require the bare-metal distributed memory installation on Echo to have a comparative base-case.
Project Timeline
- Submitted
- 01/20/2014 - 09:49