Explore a cutting-edge video presentation from ISMM 2023 that introduces a novel software framework for effective object tiering in heterogeneous memory systems. Delve into the challenges of managing multiple types of memory with varying performance characteristics in modern computing platforms. Learn how this innovative approach collects and analyzes object-level information to guide memory tiering, enabling significant performance and efficiency benefits. Discover the impact of various data tiering choices, including object prioritization for faster memory and the frequency and timing of migration events. Gain insights from results collected on a modern Intel platform featuring both conventional DRAM and non-volatile RAM, and understand how this framework compares to standard hardware- and software-directed data tiering strategies. Presented by researchers from the University of Tennessee, Intel Corporation, and Oak Ridge National Laboratory, this 18-minute video offers valuable knowledge for those interested in advanced memory management techniques and heterogeneous computing architectures.
Overview
Syllabus
[ISMM'23] Flexible and Effective Object Tiering for Heterogeneous Memory Systems
Taught by
ACM SIGPLAN