Skip to Main Content
Logical volume management over an SLC (Single-Level Cell) / MLC (Multi-Level Cell) combined flash storage can provide improved performance and reliability. This paper proposes an asymmetric flash volume management (AFVM) scheme, which exploits the asymmetry of SLC and MLC flash memory in their service time characteristics. By hot data identification and periodic data migration, AFVM tries to maximize the overall read/write performance of the logical volume. Through extensive simulations with measured data on a real smartphone, we evaluate the performance of AFVM in comparison with other placement policies. Our experiments derive the best trade-off between the performance and overheads for various values of tunable parameters such as migration unit size, migration traffic limit, and migration period. The results show that AFVM could give better solutions, significantly reducing the execution time of a trace-driven workload at the expense of a negligible decrease in flash lifetime.