We benchmark Intel Optane Persistent Memory to understand file allocation policies that optimize performance for applications that use persistent memory.
We are working on understanding large page behavior with persistent memory. We evaluate the trade-offs when persistent memory is used as storage versus memory. Understanding these trade-offs drives better decision making for file systems designed to optimize performance for applications using persistent memory.
The SNIA programming paradigm for persistent memory includes the concept of "direct access", in which the memory is mapped into the application address space, bypassing kernel file system data caching. Prior file systems (e.g, ext4, xfs, NOVA, SplitFS) focused on implementing efficient file systems, rather than the efficiency of the DAX memory provided to applications. This project explores how to construct a DAX-optimized, multi-tenant, persistent memory handling file system.
Systopia lab is supported by a number of government and industrial sources, including Cisco Systems, the Communications Security Establishment Canada, Intel Research, the National Sciences and Engineering Research Council of Canada (NSERC), Network Appliance, Office of the Privacy Commissioner of Canada, and the National Science Foundation (NSF).