| In the market of operating systems (OS) nowadays, non-open source OS, such as Windows, are occupying a big part, because of their steady performance and good after-sale services. However, the licenses of such OS are often costly, and in addition, their teams of developers are limited. Both factors restricts the performance enhancement of non-open source OS to some extent. By contrast, the open source OS have grown up rapidly in recent years. According to the Japanese Government' s statistic, in 2005, 21% of the Japanese companies and 33% of the US companies are using open source OS. In view of such a promising market of open source OS and their wide applications, the performance evaluation of these OS become more and more important. Take the example of Linux, there are various projects which are devoted to optimizing this OS. So it is important to have a clear specification for different optimizing projects, like kernel version or hardware environments.From this point of view, this article takes the example of Embedded Linux and μC/OS-II, both are open source OS, to discuss the evaluation tools of open source OS.Firstly, the principles of existing Linux evaluation tools are analyzed. Existing evaluation tools are divided into two parts. One is the BenchMark program, which runs in user mode to get performance data of the system. The other is the profile program, which inspects the kernel while the BenchMark is running. With the logged kernel data, profile can explain results of BenchMark.Then, after the analysis of profile program, it is necessary to discuss its disadvantages and then bring forward the way of optimization, such as the way to enhance the reliability of the result data or the way to reduce the rate of sample.Finally, take reference to the design principles of Linux evaluation tools. This article brings forward the design of μC/OS-II evaluation tools.In the process of the analysis and the optimization, the test result data is attached to demonstrate the point of views. |
PDF Full Text Request