CSR: Small: Cache Design for Solid-State Drives and Its Application in Data-intensive Applications

CSR：小：固态硬盘的缓存设计及其在数据密集型应用中的应用

• 批准号: 1219062
• 负责人: Miaoqing Huang
• 金额: $ 23万
• 依托单位: University of Arkansas
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2012
• 资助国家: 美国
• 起止时间: 2012-10-01 至 2015-09-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1219062&HistoricalAwards=false
• 关键词: CSR; Small; Cache; Design; Solid

Solid-State Drives (SSDs) are capable of providing much higher data access bandwidth than the hard disk drives (HDDs). However, the typical configuration of SSDs as emulating HDDs in computer systems significantly limits their impact for data-intensive applications. The objective of this research is to develop new use paradigms of SSDs in data-intensive applications on a Reconfigurable Polymorphous Processing Architecture (RPPA). RPPA consists of three major components: (1) heterogeneous processors such as CPUs and custom processing units, (2) data storage devices such as SSDs and DRAMs, and (3) a reconfigurable interconnect. The interconnect can be adaptively reconfigured as desired to fit the data communication and access pattern between the processors and the data storage devices to maximize the performance for various applications such as image processing and image registration. The direct output of this research has two folds. First, novel and application-specific use paradigms of SSD for various data-intensive applications are demonstrated. Second, the methodology and the infrastructure to develop the RPPA for exploiting the performance potentials of SSDs are established. The results of this project can transform the way SSDs are used in future computer systems for data-intensive applications. The integration of research results into several related CSE courses, such as computer organization and embedded system, provides hands-on experience in application-specific platform design to undergraduates, especially minority and women students.

固态驱动器（SSD）能够提供比硬盘驱动器（HDD）高得多的数据访问带宽。然而，SSD在计算机系统中模拟HDD的典型配置显著限制了它们对数据密集型应用的影响。本研究的目的是在可重构多态处理架构（RPPA）的数据密集型应用程序中开发新的SSD使用范例。 RPPA由三个主要组件组成：（1）异构处理器，如CPU和定制处理单元，（2）数据存储设备，如SSD和DRAM，以及（3）可重新配置的互连。互连可以根据需要自适应地重新配置，以适应处理器和数据存储设备之间的数据通信和访问模式，从而最大化各种应用（例如图像处理和图像配准）的性能。本研究的直接产出有两个方面。首先，展示了SSD用于各种数据密集型应用的新颖和特定于应用的使用范例。其次，建立了开发RPPA以开发SSD性能潜力的方法和基础设施。 该项目的成果可以改变SSD在未来计算机系统中用于数据密集型应用的方式。将研究成果整合到计算机组成和嵌入式系统等相关CSE课程中，为本科生，特别是少数民族和女性学生提供了特定应用平台设计的实践经验。