Research On Load-based Testing Of ReWorks Embedded Real-time Operating System Kernel

With the continuous progress of development technology,test verification methods and means need to be improved.Common load stress testing has been widely used in server,Web application,various information systems,network systems and so on.However,in the field of embedded real-time operating system,the research on load stress testing has not been widely and thoroughly carried out.Reworks embedded real-time operating system is independently developed by the 32 nd Research Institute of China Electronics Technology Group Corporation.It has been widely used in rail transit,information system,industrial machinery and other industries,and adopted by various equipment in the field of national defense and military affairs.It's a typical representative of equipment localization.But its testing work is basically carried out in an ideal environment without considering what the operating system may face in the actual application.Various loads and pressures can not reflect the real performance of the operating system in the application scenario,nor can they reflect its execution ability in the actual working conditions.In the use of the core processor system for a certain type of armored equipment in the army,some malfunctions such as memory overflow and system crash.At present,there is no mature embedded real-time operating system load testing tool.This thesis uses manual testing to set the load and carry out the ReWorks kernel testing research.According to the various loads which ReWorks may face in the actual operation process,four kinds of loads such as CPU load,I/O load,memory load and network load are built.Moreover,this thesis studies the construction of load test environment and load parameter setting,constructed two test environments of single load and mixed load,and proposed the load rate setting method of each load model.The proposed four load models ensure that the load randomness is more than 90% due to the random number of threads,which fills the gap of reworks load test model.Then the basic functions of ReWorks task,memory,interrupt and clock timer under the condition of single load is tested.Results show that the load has little effect on the kernel function testing,but there will be some delay in the function implementation.In order to quantitatively analyze the impact of load,four kernel performance indexes are selected,which are task switching,task preemption,interrupt response and interrupt recovery,which are related to the real-time performance of the system,In addition,this thesis proposes a series of test methods,which are tested under a variety of single load conditions.From the test results it can be concluded that after the introduction of load,the performance indicators of the system have declined,and the impact of different loads on the performance indicators of the system are different,among which the impact of network load is the most severe.The reason may be there are many interruptions under the network load,and a large number of interruptions can cause strong interference to the operation of the system.As we know,it is the first time to introduce a single load test method into ReWorks performance test.The proposed method provides a basis on quantitative analysis for the impact of different loads on the performance of the system,and raises a targeted means to explore the defects or weaknesses of the system.In order to make the load test closer to the reality and more fully reflect the real performance of the system when faced with multiple loads at the same time in the actual work,a mixed load test benchmark and mixed load test method suitable for ReWorks is proposed based on single load test.The function and performance of the system under this benchmark are tested and compared additionally.The standard average result is 19.46% lower than that under no load condition.Compared with other methods,the test environment of the proposed mixed load test method is closer to the real environment of the system operation,and the test results reflect the performance of the system more truly,which provides a better basis for the evaluation,selection and further optimization and improvement of the system.Finally,the thesis analyses the performance evaluation and optimization based on the load condition,and puts forward an evaluation scheme of introducing load and eliminating the impact of hardware.Compared to other schemes,the proposed scheme not only introduces load to make the test results closer to reality,but also avoids the impact of hardware performance degradation caused by the introduction of load,which can more accurately evaluate the performance of the system and effectively improve the performance This thesis also discusses the problem that the existing evaluation system distorts the performance evaluation of the embedded real-time operating system kernel.