The Research On Wear Leveling Algorithm For Flash Memory

Embedded devices has been developed rapidly in the research lab and used widely in many fields such as industry, military department, personal consumption. With a great widely use of embedded devices, various optimization techniques are proposed to resolve the problems of data storage and data management in these devices.With the recent technology breakthroughs in both capacity and reliability, more and more embedded systems now deploy flash memory as their storage systems, due to its small bulk, big capacity, low cost and nonvolatile features. Flash memory has gradually replaced the use of other semiconductor storages and has become the primary data and program storages. Due to the hardware architecture, data on the flash memory is written in a page and erasing is done in a block. A block must be erased before any new data is written in it. But a block could tolerate a limited number of erasing before the block becomes unreliable. The number is typically 100K to 1000K cycles. Therefore the wear leveling algorithm is to have an even erasure count distribution of flash memory blocks so that the endurance of flash memory could be improved.This paper presents the current wear leveling algorithms published in papers or patents and compares with each other in computer simulations. Then the paper lays out and prove a low bound and a upper bound of erasing count before any block reaches the maximum count. Next, two wear leveling algorithms are proposed in order to solve two different problems:During the initialization process of flash memory, block erasure information is to load into the main memory which is used to realize the even erasure. But the capacity of main memory in embedded device is not enough, reducing the consumption of main memory should be taken into account. By combining the deterministic and randomized algorithms, we invent an effective algorithm to evenly leverage the placements of cold-hot data while consumes less memory.Energy-conservation is one of the critical issues in the design of embedded systems. One promising approach is to reduce the energy consumption of secondary storage, here the flash-memory in particular. We presents a set of energy-aware flash memory management strategies. the proposed strategies can optimize the energy distribution, and the saved energy from shutting down the idle banks can effectively serve the read and write requests of applications while blocks have close erasures.