Research On The Write Efficiency And Energy Optimization Of Solid State Drives

Solid state drive(SSD) is constructed with flash chips as a block storage device. Without removable mechanic part, it has enormous advantages on read-write performance, power consumption, size and etc.. Because most of solid state drives depend on NAND-based flash memory structure, the writing of SDD needs to accomplish the block erase operation. Thus, the speed difference between reading and writing for solid state drive is large and efficiencies of them are non-symmetric.In order to improve the writing efficiency for solid state drive, it needs to upgrade the erasing operation by optimizing the mechanism of writing operation. To accomplish this, there are two feasible approaches. One is reducing the amount of the effective page copying during the operation; the other is lowering the amount of the wipe out operation in the renewal process. The two approaches are used as our solutions, the FTL integrated algorism based on the accessing frequency and the design of the SRAM flash memory transfer layer inside block.In the FTL integrated algorism based on the accessing frequency, a page identifier-"active" degree is built based on the frequency of accessing each renewal page. Then, depending on different "active" degree, all of the renewal pages will be integrated into different data blocks. By this processing, the control screening to the hot page and the hot block in solid state drive is completed. The "hot" page is the one with high renewal frequency, this kind of page has higher amount of invalid copies in the solid state drive as well. The two screenings are completed for all of the pages through the "active" degree and the "hot" area threshold by the FTL integration based on the accessing frequency, determining the hot page and the hot block, and placing the hot page and the non-hot page satisfied with double-peak distribution. As soon as solid state drive triggers the garbage collection operation, the block with the highest active degree in the hot area will be collected first, because of the higher invalid ratio in the block, the wipe out process reduces the amount of pages needed duplicated. In order to implement this algorism, the algorism's mapping mechanism uses the triad method because the mapping data saves in the flash memory array, deposited at the base and the link area respectively. The real needed data located in the flash memory array's data area. Through the analysis and deduction of the mathematical model, the restraint relation between the number of area G and the number of block in the area N is from the result of the simulation experiment, and the definition of the cleaning efficiency we can know that the cleaning efficiency is in inverse proposition to the overall performance of the solid state drive. Thus it gives a restraint condition for using the algorism; it needs to weight the cost of the dividing and grouping of the pages.In the on-drive SRAM-based FTL design, SRAM with large capacity is introduced as the writing buffer of the solid state drive. The design utilizes the localization principle of data accessing to buffer the most renewal page, thus reduce the non-place update operation in the flash memory array. With this design, the frequently page renewals are completed in the writing buffer in most of cases, because the SRAM renewal does not need erasure, it performs the efficiency of writing operation in solid state drive. In the mean time, we introduced an optimized algorism in the design of the combined operation-renewal migrating technology; further improve the efficiency of the writing operation in the certain conditions. Through deducting continuous erase operations by the mathematical model, we have got the writing model to satisfy the continuous renewal maximum n0 under the assumption of the page renewal increasing rate is in approximately fixed proportion, used integrated operation and updated migration technologies to optimize solid state drive, then finally got the consistent result with the prediction in the simulator Frey.As it is targeted on continuously optimizing the performance of solid state drive, the burden and energy consumption of the solid state drive are getting more obvious. Requirement for lowering power loss on performance optimization becomes an obstacle, so the energy saving optimization is also becoming an issue to be fixed. The energy saving technology based on Hamilton Path Re-order provides an approach, considered in the frame of I/O sub-system, for reducing the energy consumption of the solid state drive. Cjamdralacan etc. provide a theoretical base for the success of the approach. He has proved that the power consuming between the voltage switch occupies most part of the overall power consumption in solid state drive, and the switches between voltage usually occur in the process of multiple-piece solid state drive accessing individual part. Therefore, how to reduce the number of the voltage switch between different pieces of the solid state drive becomes the key of the energy saving. Thus in the algorism the issue is transferred to a access request non-vector chart that the switch number is considered as weight, and it is achieved to determine a minim weight Hamilton Path ultimately. Solid state drive applies the re-order technologies based on Hamilton Path, so that it can be sorted to the access requests provided in accordance with Hamilton Path. Therefore it can reduce the number of the voltage switch for the access requests, and depress the overall energy consumption for solid state drive.