Application-specific software defined memory
Abstract/Contents
- Abstract
- In the past three decades, the clock frequency of CPU processors has increased more than 10x. With the multi-core architecture, CPU processors continue to deliver more and more compute power each year. In contrast, the memory speed is limited by physical constraints and increasingly complicated system topology. Non-uniform memory access, or NUMA, in which one or more types of memory devices are distributed in the system, has become a new normal in today's data center. Non-uniform memory access effects often limit the performance of modern NUMA systems. Remote memory accesses have lower bandwidth and can be hundreds of nanoseconds to microseconds slower than local DRAM accesses. While many memory-intensive applications demonstrate diversified access patterns in both time and space, existing memory systems are hardwired and cannot optimize to application-specific memory access patterns. In this thesis, I will propose Software Defined Memory architecture (SDM), a novel memory system that allows the software to define application-specific policies for data placement and coherence for the critical subset of the data in a NUMA system. I will demonstrate the hardware design of SDM, and discuss the policy designs for application-specific data placement, data coherence, and software-defined memory access. The SDM design significantly improves the performance of HPC applications in a NUMA system, by 2.13x on average for investigated HPC applications, and generates 2.6x performance per dollar for critical database applications.
Description
Type of resource | text |
---|---|
Form | electronic resource; remote; computer; online resource |
Extent | 1 online resource. |
Place | California |
Place | [Stanford, California] |
Publisher | [Stanford University] |
Copyright date | 2022; ©2022 |
Publication date | 2022; 2022 |
Issuance | monographic |
Language | English |
Creators/Contributors
Author | Zhu, Chenzhuo |
---|---|
Degree supervisor | Kozyrakis, Christoforos, 1974- |
Thesis advisor | Kozyrakis, Christoforos, 1974- |
Thesis advisor | Dally, William J |
Thesis advisor | Horowitz, Mark (Mark Alan) |
Degree committee member | Dally, William J |
Degree committee member | Horowitz, Mark (Mark Alan) |
Associated with | Stanford University, Department of Electrical Engineering |
Subjects
Genre | Theses |
---|---|
Genre | Text |
Bibliographic information
Statement of responsibility | Chenzhuo Zhu. |
---|---|
Note | Submitted to the Department of Electrical Engineering. |
Thesis | Thesis Ph.D. Stanford University 2022. |
Location | https://purl.stanford.edu/wk347pj3342 |
Access conditions
- Copyright
- © 2022 by Chenzhuo Zhu
- License
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...