“NAND flash-based SSDs are widely used in data centers for their high performance and low power consumption. However, garbage collection (GC) operations are necessary due to the physical characteristics of flash memory, which can lead to latency overhead and path conflicts. The Venice scheme provides path diversity but lacks priority differentiation. A new approach, GC bypass, leverages Venice’s path diversity and enforces GC request transmission through dedicated controllers. This reduces long-tail latency by up to 25% compared to Venice, effectively mitigating interference between critical I/O operations and background maintenance tasks.”