Garbage Collection Algorithms in Flash Based Solid State Drives

Abstract

Solid state drives (SSDs)have emerged as faster and more reliable data storages over the last few years. Their intrinsic characteristics prove them to be more efficient as compared to other traditional storage media such as the Hard Disk Drives (HDDs). Issues such as write amplification, however, degrade the performance and lifespan of an SSD. This issue is in turn handled by the Garbage Collection (GC) algorithms that are put in place to supply free blocks for serving the writes being made to the flash-based SSDs and thus reduce the need of extra unnecessary writes. The LRU/FIFO, Greedy, Windowed Greedy and D choices algorithms have been described to lower write amplification for incoming writes which are different in nature. The performance of the GC algorithms varies based on factors such as pre-defined hot/cold data separation, hotness of data, uniform/non-uniform nature of incoming writes, the GC window size and the number of pages in each block of the flash memory package. Finally, it can be seen that the number of write frontiers so used, can dictate the separation of hot/cold data and increase the performance of a GC algorithm.