MFGC: Minimal First Garbage Collection for Monitoring of Context-Aware Victim Selection

Abstract

SSDs based on NAND flash are becoming a feasible secondary storage option for many computer systems. Because NAND flash memory is dif- ferent from regular HDDs, flash-based SSDs frequently have an interface layer known as Flash Translation Layer (FTL). Garbage collection tech- nique is required to free up space. Garbage collection policy is primarily concerned with cleaning costs and wear-leveling. To achieve the above design principles, we have proposed an effective minimal first garbage col- lection termed as MFGC. MFGC manages the blocks in the list, sorted in descending order of erase count. This list is divided into two regions: preference region and alternate region. The adaptive window for victim selection is used, which adapts dynamically as the cleaning cost changes. After the victim selection, effective data identification of valid hot page and valid cold page is done based on threshold and redistribution of them is done to free blocks. A Series of trace simulations is performed to evaluate the performance of proposed approach with other existing approaches. MFGC outperforms GR, CB, CAT, CATA and MCSGC by 85.5%, 77.1%,61.7%, 56.7% and 43.5% respectively.The outcome of the simulation demonstrates that our approach is superior to others in terms of erase and copy operations as well as wear leveling of blocks