Event Date
Modernizing Memory Management for Datacenter Applications
Bio
Amanda Raybuck is a PhD candidate in the department of Computer Science at the University of Texas at Austin, advised by Simon Peter. She received her BS in Computer Science from the University of Nevada, Las Vegas in 2017. She is interested in operating systems, particularly in the management of memory in the face of new application trends and memory technologies such as NVM and disaggregated memory.
Abstract
Datacenter applications continue to demand more memory. To help meet these demands, new hardware has come available. Memory technologies such as non-volatile memory (NVM), High Bandwidth Memory (HBM), and Compute eXpress Link (CXL) can help address the memory needs of applications. However, operating systems have not kept up with these advances. In this talk, I discuss how revisiting memory management techniques can help bridge this gap. First, I discuss the design of scalable tiered memory management mechanisms for datacenter applications. Then I discuss an example of a tiered memory policy for colocated datacenter applications . Finally, I look to the future and discuss ways to further advance memory management techniques for future complex memory hierarchies.