ORCID Identifier(s)

0009-0009-7303-3595

Graduation Semester and Year

Spring 2025

Language

English

Document Type

Dissertation

Degree Name

Doctor of Philosophy in Computer Engineering

Department

Computer Science and Engineering

First Advisor

Jia Rao

Second Advisor

Hong Jiang

Third Advisor

Song Jiang

Fourth Advisor

Hui Lu

Abstract

The rapid advancement of memory technologies presents new challenges and opportunities for system software and architectural design. This dissertation investigates how to optimize modern computing systems for next-generation memory, mostly focusing on persistent memory and Compute Express Link (CXL)-based memory. First, we conduct a detailed characterization of Intel Optane DC Persistent Memory, identifying the distinct behaviors of its on-DIMM read and write buffers and analyzing their impact on application performance. These insights motivate optimizations that decouple read and write paths, revealing that random read latency—especially in pointer-chasing workloads—is a dominant performance bottleneck. Second, we present NOMAD, a page management framework that enables non-exclusive memory tiering, allowing a page to temporarily exist in both fast and slow memory. NOMAD introduces transactional page migration to remove migration from the critical path and abort in-progress migrations when pages are dirtied. This design mitigates memory thrashing and improves efficiency under memory pressure. Implemented in Linux, NOMAD achieves up to 6× performance improvement over the default Linux memory manager and outperforms recent hardware-assisted tiering approaches. Finally, we propose FileShim, a POSIX-like file caching framework that facilitates efficient and consistent file sharing across isolated environments using a shared memory window and software-managed coherence. FileShim delivers up to 3× better write throughput than NFS, supports lock-free concurrent access, and makes strong file-level consistency in containerized environments. Through this work, we identify inefficiencies in current memory management approaches and highlight the need for further research to develop efficient, scalable strategies that can fully leverage emerging memory technologies.

Keywords

Disaggregated memory, Memory management, Distributed System, Operating System

Disciplines

Computer and Systems Architecture | Data Storage Systems | Hardware Systems

License

Creative Commons Attribution 4.0 International License
This work is licensed under a Creative Commons Attribution 4.0 International License.

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.