Optimizing Host-level Flash Translation Layer with Considering Storage Stack of Host Systems

Many researchers have studied and proposed various kinds of host-level flash translation layer (FTL). Host-level FTL allows the host system to handle the internal architecture of solid state drives (SSDs), and thus tries to overcome the performance or the functionality failure of traditional SSDs which only handle the I/O requests with the block I/O interface. However, existing studies only focused on the functionalities of the host-level FTL while less researches have been done on how the FTL should work on the operating system (OS) rather than bare-metal firmware. From the observation, we found that existing host-level FTL suffers from the performance bottleneck caused by unscalable software design of FTL. Therefore, we propose an optimizing scheme which efficiently processes the I/O operations requested by multiple users and guarantees the scalability of the storage stack. Our experimental results show that the performance of our scheme improves the performance of software stack in twice or more compared to the existing host-level FTL.