CSR: Small: High-Performance and Energy-Efficient Single-Level Stores: Efficient Coordinated Management of Storage and Memory

现代计算系统需要操纵持久性数据。然而，在传统的两级存储模型中，访问持久性数据的机制与操作此类数据的机制是分离的：持久性数据存储在非易失性存储器中，例如硬盘或固态磁盘，而它被操纵在非易失性存储器中，例如DRAM。 不幸的是，这种解耦的系统设计模型导致了大量的低效率，无论是在性能和能源方面。这项研究进行了重新思考的内存和存储系统中存在的新的和新兴的非易失性存储器技术，如相变存储器，自旋转移矩RAM，和忆阻器。这些技术承诺在低延迟的大容量，使他们成为一个单一的水平协调内存和存储系统的设计，其中所有数据的操作和存储是统一的主要候选人。然而，在利用新兴技术设计这样的单级存储时，存在显著的应用级和系统级效率和正确性挑战。本项目的核心目标是开发软硬件协同技术，以实现针对云计算、客户端系统和移动系统中广泛应用的计算机系统中的高性能和节能的单级存储。预计本项目的研究将实现新的应用和更强大、更高效的数据存储系统，使我们日益依赖数据处理的日常生活变得更好、更有效率。为现代计算系统启用统一的内存和存储可以大大减少浪费的工作，并利用新兴的低功耗技术，从而在使计算更节能方面迈出一大步。

Modern computing systems require manipulation of persistentdata. However, the mechanisms to access persistent data are separatedfrom the mechanisms to manipulate such data in a traditional two-levelstorage model: persistent data is stored in non-volatile storage, suchas hard disks or solid-state disks, whereas it is manipulated involatile memory, such as DRAM. Unfortunately, this decoupled model ofsystem design leads to large amounts of inefficiency in terms of bothperformance and energy.This research undertakes a rethinking of the memory and storagesystems in the presence of new and emerging non-volatile memorytechnologies such as phase-change memory , spin-transfer torque RAM,and memristors. These technologies promise large capacities at lowlatency, making them prime candidates for the design of a single-levelcoordinated memory and storage system where the manipulation andstorage of all data are unified. Yet, there are significantapplication-level and system-level efficiency and correctnesschallenges in devising such a single-level store with emergingtechnologies. The central goal of this project is to develophardware/software cooperative techniques to enable high-performanceand energy-efficient single-level stores in computer systems targetinga wide range of applications in cloud computing, client systems, andmobile systems.It is expected that research performed in this project will enable newapplications and much more robust and efficient data storage systems,making our daily lives that increasingly depend on data processingbetter and more productive. Enabling unified memory and storage formodern computing systems can largely reduce wasted work and takeadvantage of emerging low-power technologies, thereby taking a largestep in making computing more energy efficient.