Font Size: a A A

Research On Software Fault Tolerance Technology For Complex Environment

Posted on:2021-01-03Degree:MasterType:Thesis
Country:ChinaCandidate:C LiuFull Text:PDF
GTID:2518306479453984Subject:Master of Engineering
Abstract/Summary:
With the development of the computer industy,the complexity of chips is increasing exponentially.The core components of chips are becoming increasingly sensitive to environmental factors.Computers are easily disturbed by factors such as space radiation,circuit fluctuation,which always lead to system failure.The data flow error and control flow error cause by soft error are the main manifestation.Therefore,the research has important practical significance and application value in software reliability.This thesis mainly studies the software fault tolerance technology for complex environment,including the following.We propse a data flow error detection method based on deep forest to address the problem that traditional methods have low detection rate and high overhead.First,we employ the program analysis to extract instruction features which are highly related to SDCs.Second,we design the deep forest regression model to predict the SDC vulnerability of instructions.Third,we optimize the error detection procedure by duplicating the critical instructions with different granularity.Finally,we evaluate our DFRMR model on Mibench benchmarks with multiple testing programs.The results show that our method attains better detection accuracy compared to other state-of-the-art methods and keeps the low multi-granularity redundancy.We propose a control flow error detection method based on bipartite block to address the problem that traditional methods have low error coverage.First,we analysis the transfer pathway of control flow to design the strategy of block classification and signature assignment.Second,we design a method to detect the control flow error intra block based on bipartite gragh.Finally,we evaluate our method on benchmarks with multiple testion programs.The results show that our method attains high control flow error detection rate and low overhead without intra block fault tolerance,and attains higher detection rate if we employ the intra block code instrumentation.We implement the fault tolerance system based on LLVM、LLFI etc.Programs that are processed by the system can get the ability of data flow and control flow error detection.The experiment results show that the system can effectively protect the target programs.
Keywords/Search Tags:Soft error, SEU, data flow error detection, control flow error detection, deep forest
Related items