Integrating SSS into Storage Hierarchy

Flash memory is increasingly adopted as main data storage media in mobile devices, notebooks, desktops, and enterprise class servers. This is mainly due to the flash memory superior characteristics, over conventional disk drives that includes smaller size, lighter weight, lower power consumption, shock resistance, lesser noise, non-volatility memory, and faster read performance. Replacing magnetic convectional disks by solid state disks (SSD) would result in significantly improved performance. However, there are still significant opportunities to exploit distinguishing SSD’s characteristics in a wide variety of applications. In this project, we aim to explore several new approaches to exploiting SSD characteristics that will significantly enhance the performance of selected applications including file systems and database systems.

Status

Currently we focus on: (1) Caching where we aim to develop a new caching algorithm for SSD, which considers asymmetric read and write performance of flash memory to make cache management decisions. (2) Spatial Indexing (for example, R-Tree) where we design a special and efficient implementation for various spatial indexes running on SSDs. (3) Frequent data updates where we improve write performance and longevity of flash based SSDs by designing an efficient scheme for handling data updates. (4) Checkpointing where we propose a new architecture for checkpoint storage system that uses SSD as an intermediate medium for storing the checkpointing data before saving it on the stable storage.

Publications

Last modified 23 May 2019