Optimization Mechanism Of Wear-Out And Workload Balancing Based On Page Switching

Solid-state drives(SSDs)based on flash memory chips have become the mainstream storage devices due to their superior performance compared to traditional Hard Disk Drives(HDDs).However,the flash memory chip can only withstand a limited number of erase operations.Excessive erase operations will bring a lot of wear-out,which will not only deteriorate the data reliability of SSDs,but also reduce the lifespan of SSDs.Therefore,RAID is applied to SSDs to improve their system-level reliability,a parity chunk is maintained for each stripe to ensure data reliability.When a SSD is damaged and data is lost,the RAID mechanism can recover the lost data in a timely and effective manner.And when updating data on a RAID-based SSDs,not only the data bit but also the parity bit of the same stripe need to be updated,which will generate a large number of parity bit updates,and the new data must be written to the channel where the original data is located.This can lead to an unbalanced wear-out and I/O workload degree between the channels of the SSD,reducing its lifespan and I/O performance.Therefore,how to improve the SSDs in each channel balance is particularly important.In view of the above problems,this thesis completed the following work:Firstly,evaluate the imbalancing degree of the SSDs,and verified that some data bits would generate more update operations than parity bits,which provides the motivation for this thesis.The mathematical model presented in this paper evaluates the degree of balance between channels in SSDs.By considering a number of relevant indicators,the model can be used to judge whether the internal channels of the SSD are in a balanced state,and quantify the unbalance degree of the system,which provides guidance and basis for data exchange operation.Secondly,in view of the phenomenon of internal wear-out and I/O workloads imbalance in SSD,this thesis proposes a workload and wear-out balancing optimization mechanism based on page switching.When a data update request arrives,the access heat of the data bit and parity bit,as well as the degree of wear-out and workload between each channel,are fully considered to select the appropriate data exchange mode.Through evaluation of the established mathematical model,data exchange operations can be performed when the balancing degree between the channels of the SSDs is improved.Through comprehensive consideration of the condition of each channel and the data itself,part of the wear-out and workload in the channel with heavy wear-out or workload is switched to the channel with the lighter degree,so as to achieve wear-out and workload balance,as well as to improve the lifespan of SSDs and reduce the system I/O response time.Finally,in order to verify the effectiveness of the proposed optimization mechanism,this thesis obtained experimental results of this mechanism using a real dataset in the SSDSim simulator and compared it with related work.The experimental results show that the proposed mechanism not only effectively improves the workload balancing between the channels inside the SSDs and reduces the system’s I/O response time,but also makes the wear-out degree between the channels more balanced,which extends the lifespan of the SSDs.