【摘要】：性能分析對(duì)構(gòu)建高效率的計(jì)算系統(tǒng)意義重大,但隨著云計(jì)算的流行,計(jì)算系統(tǒng)日益復(fù)雜,性能分析工作變得十分困難。而現(xiàn)有的性能分析工具,主要采取用戶行為模擬等手段,關(guān)注用戶體驗(yàn),監(jiān)測(cè)響應(yīng)時(shí)間、錯(cuò)誤率等參數(shù),弱化了分析部分;同時(shí)只適用于Web類用戶交互應(yīng)用,無(wú)法應(yīng)用于諸如科學(xué)計(jì)算等領(lǐng)域。因此,下一代性能分析工具的設(shè)計(jì)開發(fā)工作迫在眉睫。本文基于機(jī)器學(xué)習(xí)算法設(shè)計(jì)了一種深入挖掘應(yīng)用各項(xiàng)性能指標(biāo)內(nèi)在聯(lián)系,統(tǒng)籌多指標(biāo)聯(lián)合定位應(yīng)用問(wèn)題的專家決策系統(tǒng),解決了現(xiàn)有分析工具定位分析問(wèn)題能力不足,適用場(chǎng)景有限的問(wèn)題。本文首先介紹了背景知識(shí)及開發(fā)實(shí)現(xiàn)所涉及的相關(guān)技術(shù),并調(diào)研和說(shuō)明了現(xiàn)有分析工具的功能和特點(diǎn)。通過(guò)分析專家決策系統(tǒng)在整個(gè)性能評(píng)估系統(tǒng)中的位置,結(jié)合RTC應(yīng)用分析實(shí)例,得出決策系統(tǒng)的工作場(chǎng)景和總體需求。然后,提出了專家決策系統(tǒng)的總體結(jié)構(gòu),并對(duì)、決策系統(tǒng)的關(guān)鍵流程進(jìn)行了闡述,并給出了決策系統(tǒng)的內(nèi)外接口。在此基礎(chǔ)上,設(shè)計(jì)了專家決策系統(tǒng)的類圖,并針對(duì)各個(gè)模塊中涉及的關(guān)鍵問(wèn)題,給出了詳細(xì)的流程設(shè)計(jì),完成了決策系統(tǒng)原型的設(shè)計(jì)和實(shí)現(xiàn)工作。隨后對(duì)專家決策系統(tǒng)進(jìn)行了功能測(cè)試和驗(yàn)證,詳細(xì)的介紹了關(guān)鍵功能模塊的測(cè)試用例,測(cè)試結(jié)果表明功能原型基本符合需求。最后,對(duì)專家決策系統(tǒng)的開發(fā)設(shè)計(jì)工作進(jìn)行了總結(jié),并指出了下一步的幾個(gè)主要研究方向。
[Abstract]:Performance analysis is of great significance to the construction of high-efficiency computing system, but with the popularity of cloud computing, the computing system is becoming more and more complex, and the performance analysis work becomes very difficult. And the existing performance analysis tool mainly adopts the means of user behavior simulation, so as to pay attention to the parameters such as user experience, monitoring response time, error rate and the like, weaken the analysis part, and can only be applied to the interactive application of the Web user, and can not be applied to the fields such as scientific calculation and the like. Therefore, the design and development of the next-generation performance analysis tool is urgent. This paper, based on the machine learning algorithm, designs an expert decision-making system for deep excavation and application of various performance indexes and integrated multi-index joint positioning and application, and solves the problem that the existing analysis tool's ability to locate and analyze the problem is not enough and the scene is limited. This paper first introduces the background knowledge and the related technology involved in the development, and studies and explains the functions and characteristics of the existing analysis tools. By analyzing the position of the expert decision-making system in the whole performance evaluation system, combined with the analysis example of the RTC application, the working scene and the overall demand of the decision-making system are obtained. Then, the overall structure of the expert decision-making system is put forward, and the key process of the decision-making system is described, and the internal and external interfaces of the decision-making system are given. On this basis, the class diagram of the expert decision-making system is designed, and the detailed process design is given for the key problems involved in each module, and the design and implementation of the prototype of the decision-making system are completed. Then, the function test and verification of the expert decision-making system are carried out, and the test cases of the key function modules are described in detail. The test results show that the functional prototype basically conforms to the requirements. Finally, the development and design of the expert decision-making system are summarized, and the main research directions of the next step are pointed out.
【學(xué)位授予單位】：北京郵電大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2016
【分類號(hào)】：TP182;TP311.52

【參考文獻(xiàn)】

相關(guān)期刊論文 前1條

1 吳江;黃晟青;蔡駿;;互聯(lián)網(wǎng)購(gòu)物網(wǎng)站用戶體驗(yàn)設(shè)計(jì)研究[J];包裝工程;2012年08期

，

本文編號(hào)：2506929