CSR: Small: Prediction-Based Data Placement for New Memory and Storage Hierarchies

点击翻译按钮获取中文摘要

Many high-end computing and data-intensive applications, such as transaction processing systems, large-scale scientific simulations, and on-demand services, are limited by the performance of the underlying memory and storage systems. The goal of this project is to improve the memory/storage hierarchy by integrating new and emerging memory/storage technologies with the existing memory/storage hierarchy. Flash memory based Solid State Drives (SSDs) and Phase Change Memory (PCM) are two emerging memory/storage technologies that show tremendous promise to improve the performance of current computer systems. Shingled Write Disks (SWDs) have been recently proposed to further improve the magnetic disk space capacity to another level. This project is to investigate effective data placement in the new memory/storage hierarchies to better support high-end computing and data-intensive applications. Considering the special characteristics of these new technologies, this project provides innovative solutions for prediction-based data placement in the new memory/storage hierarchies to effectively improve the performance of both memory and storage data accesses. It speeds up the integration of PCM with DRAM as a new memory hierarchy and flash memory-based SSDs and SWDs with the existing storage hierarchy. Providing efficient solutions to the fundamental performance issues greatly improves the performance of many critical applications including weather forecasting, large-scale scientific simulations, e-commerce, etc. Therefore, the project makes huge impact on our daily life. Quickly integrating SWDs into the existing storage hierarchy will also enhance our capability to handle huge amounts of available data. This project creates positive impact on every layer of the memory and storage hierarchies.