Red-black tree I/O management of solid state disk with elastic striping design.

With the rise of big data, 1 the demand for hard disk performance is also getting higher and higher. Because traditional hard drives are gradually being replaced by solid-state hard drives, The shortcomings of the solid state hard disk itself have gradually surfaced. SSD also have a number of disadvantages: Write is slower than read, limited lifetime, erase-before-write, and write and read cannot be done at the same time, How to avoid these shortcomings leads to slower performance, And how to improve it will be a hot topic that has attracted much attention. In this paper, we have improved the system's I/O scheduling and hard disk writes. The I/O scheduling of the original system could not take advantage of the solid-state drive itself, so we improved the original I/O scheduling and added a red-black tree algorithm to make each requested schedule have certain fairness. In addition, when storing data on a solid state drive, the original features erase old data that needs to be overwritten before writing new data. To avoid the need to wait for the erasing time every time new data is written, the characteristics of the elastic block are used to write each new data directly into the new magnetic block. Only then will the old data space be released. This method allows new data to be overwritten without having to wait for the operation time to erase the old data before the new data can be written. Therefore, the response time when data is overwritten is greatly reduced. And with the collation of the flash memory controller, the cold data and the hot data are written into the cold block and the hot block respectively. To avoid write amplification and improve the problem of uneven loss, garbage collection also uses the background garbage collection method to wait for the system to be garbage collected when it is idle, which also greatly improves the overall system performance.