CSR:Small:Supporting Position Independence and Reusability of Data on Byte-Addressable Non-Volatile Memory

CSR：Small：支持字节可寻址非易失性存储器上数据的位置独立性和可重用性

• 批准号: 1717425
• 负责人: Xu Liu
• 金额: $ 50万
• 依托单位: North Carolina State University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2017
• 资助国家: 美国
• 起止时间: 2017-08-15 至 2022-07-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1717425&HistoricalAwards=false
• 关键词: CSR; Small; Supporting; Position; Independence

With the potential to eliminate the boundary between memory and storage, Non Volatile Memory (NVM) calls for innovations across computing stacks. Existing studies have been mostly focused on the how to ensure data consistency and recoverability when a system crashes with minimum overhead. This proposal aims to enable effective leverage of NVM persistency for computing efficiency. In particular, the research focuses on a central challenge for data reuse: To efficiently enable position independence for dynamic data structures such that all the references in a data object keep their integrity (i.e., point to the correct addresses) regardless of where in the memory space the data object is mapped.NVM is expected to be influential for many computational fields, from database to data analytics, scientific computing, analysis of social networks, modeling of stock transactions and so on. The success of this project may provide key support for effective data reuse on NVM, helping improve computing efficiency and software development productivity on future systems, and hence sustained advancement of science, health, and security.The PIs have a version controlled repository in place at NCSU that has and will be used for data storage and access control. For metadata, documentation will be submitted alongside the data in order to make the data reusable. The data will be archived for at least 10 years.

由于有可能消除内存和存储之间的界限，非易失性内存（NVM）要求跨计算堆栈进行创新。现有的研究主要集中在如何在系统崩溃时以最小的开销保证数据的一致性和可恢复性。该提案旨在有效利用NVM持久性来提高计算效率。特别是，研究集中在数据重用的核心挑战：有效地启用动态数据结构的位置独立性，以便数据对象中的所有引用保持其完整性（即指向正确的地址），而不管数据对象在内存空间中的映射位置。预计NVM将对许多计算领域产生影响，从数据库到数据分析、科学计算、社会网络分析、股票交易建模等。该项目的成功可能为NVM上有效的数据重用提供关键支持，有助于提高未来系统的计算效率和软件开发生产力，从而持续推进科学、健康和安全。pi在NCSU有一个版本控制的存储库，它已经并将用于数据存储和访问控制。对于元数据，文档将与数据一起提交，以使数据可重用。这些数据将被存档至少10年。

项目成果

Preserving Addressability Upon GC-Triggered Data Movements on Non-Volatile Memory

• DOI: 10.1145/3511706
• 发表时间: 2022-01
• 期刊: ACM Transactions on Architecture and Code Optimization (TACO)
• 作者: Chencheng Ye;Yuanchao Xu;Xipeng Shen;Hai Jin;Xiaofei Liao;Yan Solihin

Exploring Flexible Communications for Streamlining DNN Ensemble Training Pipelines

• DOI: 10.1109/sc.2018.00067
• 发表时间: 2018-11
• 期刊: SC18: International Conference for High Performance Computing, Networking, Storage and Analysis
• 作者: Randall Pittman;Hui Guan;Xipeng Shen;Seung-Hwan Lim;R. Patton

Supporting Legacy Libraries on Non-Volatile Memory: A User-Transparent Approach

• DOI: 10.1109/isca52012.2021.00042
• 发表时间: 2021-06
• 期刊: 2021 ACM/IEEE 48th Annual International Symposium on Computer Architecture (ISCA)
• 作者: Chencheng Ye;Yuanchao Xu;Xipeng Shen;Xiaofei Liao;Hai Jin;Yan Solihin

MERR: Improving Security of Persistent Memory Objects via Efficient Memory Exposure Reduction and Randomization

MERR：通过有效减少内存暴露和随机化来提高持久内存对象的安全性

• DOI: 10.1145/3373376.3378492
• 发表时间: 2020
• 期刊: ASPLOS'20: Architectural Support for Programming Languages and Operating Systems
• 作者: Xu, Yuanchao;Solihin, Yan;Shen, Xipeng
• 通讯作者: Shen, Xipeng

Efficient Document Analytics on Compressed Data: Method, Challenges, Algorithms, Insights

压缩数据的高效文档分析：方法、挑战、算法、见解

• DOI: 10.14778/3236187.3236203
• 发表时间: 2018-07
• 期刊: PROCEEDINGS OF THE VLDB ENDOWMENT
• 影响因子: 2.5
• 作者: Zhang Feng;Zhai Jidong;Shen Xipeng;Mutlu Onur;Chen Wenguang
• 通讯作者: Chen Wenguang