Research And Implementation On Non-Volatile Memory Management Schemes

With the growth of the requirements of capacity and speed of storage and memory systems,Non-Volatile Memory(NVM)has developed rapidly in recent years.Besides non-volatile and low price,the advantages of NVM also include extremely low idle power consumption and reading speed close to that of DRAM(Dynamic Random Access Memory).These benefits attract so many scholars around the world to research on it.However,NVM also has many disadvantages.For example,writing speed of NVM is slower than that of DRAM and writing power consumption is higher than that of DRAM.Additionally,the number of NVM writing is limited.When NVM is overwritten,it will be damaged permanently.In order to apply NVM into practical systems to produce the advantages and overcome the disadvantages,this thesis does these studies as follows:1.In order to apply NVM to automotive electronic systems,this thesis first proposes a static data management scheme VSDA(Vehicle Static Data Allocation).By analyzing the common characteristic rules of data in automotive electronic systems,the driving produce is divided into three phases.Different management methods in VSDA are applied to these three phases to reduce the writing wear of the NVM systems.The experiment shows the effectiveness of VSDA on reducing the NVM writing operations when the driving time is long enough and the rules of data are steady.2.The static data management scheme needs to classify data artificially and VSDA has some limitations when driving time is short or the data rules are not stable.This thesis proposes a dynamic data management scheme VDDM(Vehicle Dynamic Data Management).By analyzing the sampling rates and change rates of data,VDDM predicts data changes roughly so that it can manage data dynamically.The experiment shows that VDDM can reduce total writes on NVM from 72.94% to 96.5% compared with HIRLIN(High sampling rates In RAM Low sampling rates In NVM)and ONA(Only NVM Architecture).VDDM can prolong the lifetime of NVM effectively.3.To avoid the premature termination of NVM lifetime from uneven writing of VDDM,this thesis optimizes VDDM scheme.This thesis proposes a VDM(Vehicle Date Management)which provides wear leveling mechanism VSWL(Vehicle-Specific Wear Leveling)and fault tolerance mechanism VFT(Vehicle Fault Tolerance)to support automotive ECU data management scheme.The experiment shows that the memory can be written more than 7.9 to 16.8 times using VDM compared with using HIRLIN and ONA.VDM can prolong the lifetime of NVM effectively.4.Besides being applied to automotive electronic systems,NVM also can be used to consumer electronics to optimize system power consumption.For smart bracelet,this thesis analyzes and models the reading and writing power consumption on NVM and DRAM and the corresponding mathematical descriptions are defined.This thesis proposes a power-aware data management scheme BPADM(Bracelet Power-Aware Data Management)for smart bracelet.The experiment shows that BPADM can reduce power consumption from 36% to 82% in normal mode compared with DMA(Dynamic Measure and Allocation),ONA and noManage methods,and BPADM can reduce power consumption from 25% to 66% in sleeping mode.BPADM can prolong the stand-by time of smart bracelet.