In solid-state drives (SSDs), data is not generally updated in place. Rather, data update in SSDs involves garbage collection, in turn facilitated by overprovisioning of storage capacity. Overprovisioning reduces actual available storage capacity. Garbage collection results in write amplification, an undesirable phenomenon where the amount of information physically written is a multiple of the logical amount of information to be written. Write amplification reduces the throughput and the life of the SSD. This disclosure describes techniques to adaptively classify data based on its coldness, e.g., recency and frequency of access, and dynamically allocate overprovisioning rates for each coldness category to achieve optimal write amplification and throughput.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Tan, Bin; Shahidi, Narges; and Benitez, Ricky, "Adaptive Data-Coldness Classification For Solid State Drives", Technical Disclosure Commons, (October 26, 2020)