Research On Write Scheduling And Data Reallocation Based On Read Disturb In Solid State Disk

With the continuous development of data storage technology,solid-state hard drives are increasingly used in scenarios such as smart phones,servers,and personal computers.With the ever-increasing application of solid-state hard drives today,various noise disturb have brought all kinds of additional burdens to the use of solid-state drives.Among the various disturb,the phenomenon of read disturb is expected to have more and more serious adverse effects as the density of flash memory cells continues to increase and read-intensive applications continue to increase.The source of read disturb is that when a single page in the flash memory block is read,high voltages are frequently applied to all flash memory pages that have not been read,which causes abnormal leakage of charges in the gate and further increases the read error rate.Problems such as longer reading time.How to optimize software-level strategies for read disturb is a hot topic in the current research on solid state drives.In response to this problem,this paper has launched a research on the mechanism of data mapping and data reorganization based on read disturb.The main research content includes the following three aspects.First,on the premise of analyzing and summarizing the influencing factors of read disturb,by analyzing the data provided in the literature and using methods such as nonlinear regression,a model to measure the read error rate of the flash memory block is established.The input factors of the model are the programming/erase times and the reading times of the flash memory block,and the output result is the read error rate of the flash memory block.Afterwards,it is verified that the proposed model has good accuracy through deduction.Secondly,after establishing a model to measure the read error rate of the flash memory block,by judging the read error rate of the flash memory block,the flash memory block is divided into two types that are susceptible to read disturb and those that are not easily affected by read disturb.Then analyze the written data request,extract the hot read data in the request,and write it to the flash memory block that is not susceptible to read disturb when the hot read data is written,and write it to the LSB first(Least Significant Bit)page.Blocks that are not easily affected by read disturb have better adaptability to program/erase operations and read operations,while LSB pages require less frequent reference voltages when being read,both of which can further reduce the impact of read interference.Then,based on the unbalanced distribution of the traditional read refresh erase,a data reorganization method based on read balance was proposed,and the traditional read refresh method was optimized.The proposed method first uses a model to measure the read error rate,and divides the flash memory block into two types that are susceptible to read disturb and that is not easily affected by read disturb.Then,when reading and refreshing,the hot and cold data are evenly distributed in the two blocks.Placed to achieve the effect of equalizing the number of internal readings of the two blocks.We also derive formulas to prove that when the read times of each flash memory block are balanced,the overall read error rate will be the lowest,which provides theoretical support for the data reorganization strategy based on read balance.Finally,in order to verify the effectiveness of the proposed optimization scheme,this paper conducts a simulation experiment on the newly proposed improvement scheme on the SSDSim simulator.The experimental results show that the newly proposed improvement scheme effectively alleviates the negative effects of increased read error rate caused by read disturb,longer read time,and more read refresh times,and effectively improves system performance.