A Novel Update Scheme for High-Performance Transaction Processing Systems with SSD Storage

With advancements in networking and semiconductor technology, the volume of transactions processed in real-time is significantly increasing.This trend is expected to accelerate further with the use of artificial intelligence for data monitoring and automated processing, especially in mobile transaction processing and intelligent online payments.Consequently, the importance of DBMSs capable of handling a large volume of transactions per unit of time has grown significantly.To meet this demand for high-performance DBMSs, there is a strong need to utilize flash memory SSDs (Solid State Drives) as the primary storage for DBMSs.This is because the SSD storage offers consistent and fast page read times due to the physical characteristics of NAND flash memory.However, SSDs suffer from high page update costs and require periodic garbage collection.In this paper, we propose a novel scheme to mitigate these drawbacks by leveraging the Redo log data stored by DBMSs for system recovery, thereby providing the avoidance of frequent page updates.Since the proposed scheme is designed based on the well-known ARIES recovery algorithm, it is easily integrated into existing DBMS architecture.The proposed scheme will enhance the processing speed of update-intensive transactions and expedite the recovery time of SSD-based DBMSs.