DNN SSD SYSTEM AND METHOD FOR OPTIMIZING PERFORMANCE OF A SOLID-STATE DRIVE USING A DEEP NEURAL NETWORK

摘要

A controller of a data storage device comprises: a host interface providing an interface for a host computer; a flash translation layer (FTL) converting a logical block address (LBA) into a physical block address (PBA) related to an I/O (input/output) request received from the host computer through the host interface; a flash interface providing an interface for a flash media of the data storage device and accessing data stored on the flash media of the data storage device; and one or more deep neural network (DNN) modules predicting an I/O access pattern of the host computer. The host interface provides one or more input vectors for one or more DNN modules, and the one or more DNN modules use one or more vectors to provide one or more predicted outputs for the FTL related to one or more past I/O requests and present I/O requests received from the host computer. One or more predicted outputs include at least one predicted I/O request following the present I/O request. The FTL prefetches data stored in the flash media related to at least one predicted I/O request. The SSD controller uses information related to data access patterns based on various indicators to optimize performances of SSD related to usages of a bandwidth, waiting time and internal resources.