· Contributors · Organizations · Search
Data-Centric Resource Management for Complex Memory Fabrics
System Software and Runtime Systems
DescriptionThe continued need for scaling of performance, capacity, and cost efficiency for data-intensive applications, is giving rise to new system designs combining heterogeneous memory components, accelerators and accelerator-near memory, interconnected with programmable high-performance interconnects. Maximizing the benefits of the resulting complex memory fabrics requires new methods for coordinating decisions concerning data placement, movement, accesses and transformations, across different system components and, currently, independent resource managers. This talk will discuss some of the new data-centric resource management problems that must be addressed to deliver on the technology expectations and summarize results from our group’s recent research in this space.