CAREER: Tags: A Unifying Primitive to Build Storage Data Paths for Swiftly Evolving Workloads and Storage Media

职业：标签：为快速发展的工作负载和存储介质构建存储数据路径的统一原语数字存储现在是我们生活的一部分。 然而，传统操作系统中的存储管理由于存储介质（例如，存储卡和拇指驱动器）和新的用户需求（例如，能源效率、安全性等）。 特别是，遗留存储管理假定使用磁盘而不是基于内存的存储，并且是分层结构的，跨层协调性差，对数据密集型应用程序的支持很少。一种可用于构建存储层的统一原语。 这个想法是用它们如何相互关联以及它们如何被每一层处理来标记每一段数据。 标签将简化存储服务组合和数据共享。 例如，文件搜索可以基于其与电子邮件和日历事件的关联。 标签将简化适应新存储介质所需的系统更改的隔离。 标签描述数据片段的性质（例如，机密、流媒体等）使得存储层可以协调以改进性能、能量效率等。标签将跟踪数据依赖性使得机密数据和衍生物（例如，知识价值：本研究的挑战包括：（1）探索标签的表示和接口;（2）在可伸缩性、安全性以及使用模型和存储介质的假设的约束下检查标签的管理;（3）开发通过标签跟踪数据依赖性的方法;（4）开发基于标签的数据管理。以及（4）展示标签在不同使用模式和存储介质上的益处。更广泛的影响：标签支持数据密集型应用程序，如数据库和搜索引擎。标签的设计类似于机器指令，并允许应用硬件优化技术。标签提供的数据跟踪服务将支持安全存储服务。 标签框架将成为一个有价值的教学工具，一个探索新思想的平台，一个探索其他领域思想的门户。 这项研究将导致一个关于存储的研讨会课程的发展，强调从各个领域注入思想，以推进存储前沿。

CAREER: Tags: A Unifying Primitive to Build Storage Data Paths forSwiftly Evolving Workloads and Storage MediaDigital storage is now a part of our lives. However, storage management in legacy operating systems becomes quickly outmode due to swift advances in storage media (e.g., memory cards and thumb drives) and new user requirements (e.g., energy efficiency, security, etc.). In particular, legacy storage management presumes the use of disks rather than memory-based storage and is structured in layers with poor cross-layer coordination and little support for data-intensive applications.This project introduces tags?a unifying primitive that can be used to build storage layers. The idea is to tag each piece of data with how they relate to each other and how they are handled by each layer. Tags will ease storage service composition and data sharing. For example, file searches can be based on its association with e-mails and calendar events. Tags will ease the isolation of system changes required to adapt to new storage media. Tags describe the nature of data pieces (e.g., confidential, streaming media, etc.) so that storage layers can coordinate to improve performance, energy efficiency, etc. Tags will track data dependencies so that confidential data and derivatives (e.g., temporary files) can be handled securely.Intellectual merits: The challenges for this research include: (1) exploring the representation and interface of tags; (2) examining the management of tags under constraints of scaling, security, and assumptions of usage models and storage media; (3) developing ways to track data dependencies via tags; and (4) demonstrating the benefits of tags across diverse usage patterns and storage media. Broader impacts: Tags support data-intensive applications such as databases and search engines. The design of tags is similar to machine instructions and allows the application of hardware optimization techniques. The data-tracking service offered by tags will support secure storage services. The tag framework will become a valuable instructional tool, a platform to explore new ideas, and a gateway to explore ideas from other fields. The research will lead to the development of a seminar course on storage that emphasizes on infusing ideas from various fields to advance storage frontiers.