A Group-Based Hybrid Wear-Leveling Algorithm for Flash Memory Storage Systems

Wear-leveling algorithm is required in most flash memory controllers to prolong the lifetime of flash memory by distributing the erase operations evenly across the whole flash memory. This paper proposes a group based wear-leveling algorithm which can dramatically reduce the RAM space required by wear-leveling algorithm while maintaining the wear-leveling performance. We adopt random wear-leveling policy which requires very little RAM space inside a group to make the data in the group move around and the wear in the group respectively even. And a threshold controlled static wear-leveling policy is adopted among different groups to make the average wear of different groups even. A series of trace-driven simulations show that our group based algorithm consumes only 78% of RAM space compare with previous group-based wear-leveling algorithm which require complex data structure, while maintaining nearly the same wear-leveling performance.