A quantitative framework for modeling and analyzing flash memory wear leveling algorithms

Flash memory is emerging as an enabling technology rapidly changing the landscape of storage systems via its many desirable properties such as low power consumption and high random I/O throughput. However, due to its physical characteristics, flash cells have limited erase (program) cycles. To increase endurance and reliability, flash memory based devices employ certain wear leveling algorithms. But so far there hasn´t been a rigorous mathematical tool to analyze and evaluate the effectiveness of various wear leveling algorithms. In this paper, a mathematical framework is presented to model whole block wear leveling in flash-based devices using probabilistic models of workloads, strategies and wear level states. Then equations are derived for the distribution of wear levels to enable quantitative analysis and evaluation of wear leveling algorithms for flash memory based devices.