Remap-SSD: Safely and Efficiently Exploiting SSD Address Remapping to Eliminate Duplicate Writes

Duplicate writes are prevalent in diverse storage systems, originating from data duplication, journaling, and data relocations, etc. As flash-based SSDs have been widely deployed, these writes can significantly degrade their performance and lifetime. To eliminate duplicate writes, prior studies have proposed innovative approaches that exploit the address remapping utility inside SSDs. However, remap operations lead to a mapping inconsistency problem, which may cause data loss and has not been properly addressed in existing studies. In this paper, we propose a novel SSD design, called Remap-SSD, with two notable features. First, it provides a remap primitive, which allows the host software and SSD firmware to perform logical writes of duplicate data at almost zero cost. Second, a hybrid storage architecture is employed to maintain the mapping consistency. Small byte-addressable non-volatile RAM (NVRAM) is used to persist remapping metadata in a log-structured manner and is managed synergistically with flash memory. We verify Remap-SSD on a software SSD emulator with three case studies: intra-SSD deduplication, SQLite journaling, and F2FS cleaning. Experimental results show that Remap-SSD can realize the full potential of address remapping to improve SSD performance and lifetime.