DATA MANAGEMENT IN A FLASH MEMORY BASED STORAGE SERVER

Abstract

Flash memory has many attractive features, such as nonvolatility, light weight, and low power consumption. These features show promise for using flash memory as storage in consumer electronics, embedded systems, and mobile computers. However, flash memory has specific hardware characteristies that impose challenges on the design of storage systems. It cannot be overwritten unless erased in advance. The earse operations are slow and power-wasted, which usually decrease system performance and consumed lots of power. In addition, the unmber of times that flash memory can be erased is also limited. This paper descries the design and implementation of a storage server for flash memory. To overcome hardware limitations, the server employs an effctive dynamic data clustering method and an efficient cleaning algorithm. Performance evaluation shows that, with these mechanisms, throughput is significantly improved, flash memory lifetime is prolonged, and even wearing is ensured.