Method for improving SSD comprehensive performance

Abstract

The invention discloses a method for improving the SSD comprehensive performance. The method is characterized in that two stages of caches are adopted as writing caches of a solid state disk controller, the first stage is a partial-write cache, the second stage is a full-write cache, the partial-write cache and the full-write cache are managed with the size of physical pages as the cache page unit, one physical page comprises logic data blocks of N hosts, a writing command received by the solid state disk controller from the hosts includes the steps of writing data in the partial-write cache first, transferring the data of cache pages in the partial-write cache into the full-write cache when the data of a certain cache page in the partial-write cache is fully written, a background of the solid state disk controller only executes to write the data cached in the full-write cache in an actual FLASH storage space, and the partial-write cache temporarily does not write data in the actual FLASH storage space. By arranging the two stages of caches, it is ensured that full data is actually written in each page of the FLASH storage space, the triggering probability of read-modify write (RMW) is greatly reduced, and the SSD comprehensive performance is remarkably improved.

Description

Technical field [0001] The present invention relates to solid state hard disk control technology, in particular to a method for improving the comprehensive performance of SSD. Background technique [0002] For enterprise-class solid state drives (SSD), it is not comprehensive to consider SSDs only by performance and IOPS (Input / Output Operations Per Second, that is, the number of read and write (I / O) operations per second). More important is the quality of service, which is Qos (quality of service). [0003] Service quality can be simply understood as the consistency or stability of SSD performance. It is not simply bandwidth, IOPS or average latency, but the average performance of the SSD after these factors are combined. In an enterprise-level application environment, every second of delay may have a great impact on the business. For enterprise-level data centers that require everything in their hands, the consistency of SSD performance is particularly important. [0004] figure...

AI Technical Summary

Problems solved by technology

When the host continues to initiate random services at the LBA level, the performance of the SSD will be extremely unstable and the service quality will be poor

Images