嵌入式網路通訊裝置評比技術與工具之研發---子計畫三：嵌入式網路通訊裝置核心與網路協定行為效能分析評比技術與工具之研發(中心分項)(I)

Abstract

本子計畫為期二年，延續執行中的計畫成果，本子計畫主要目標是發展嵌入式網路通訊裝置核心與網路協定行為效能分析評比之技術與工具，以協助嵌入式裝置開發者診斷通訊行為的缺點和瓶頸，縮短研發週期。 嵌入式網路通訊裝置的通訊行為除了基本的網路存取(access)能力外，我們經常關心的是傳輸延遲(delay)、反應時間(response time)以及封包遺漏(packet loss)等效能指標。這些效能指標會因為通訊行為中的(1)系統內部的(internal)核心行為，如：作業系統(operating system)核心內部網路相關元件(component)的行為、元件間的訊息交換(signaling)等，以及(2)外顯的(external)網路協定(protocols)行為，如：DHCP、TCP、IP、IEEE802.11等的影響而有所變化。但目前仍沒有一套合適的工具能協助開發者，分析這些指標以及造成延遲與封包遺漏的原因。 現有的網路封包(packet)擷取工具，僅能協助分析外顯的網路協定行為；目前的核心分析工具也只能追蹤函式執行順序、執行時間以及指定的記憶體內容，但並未能追蹤核心事件(例如：Link-up、Link-down、IP depletion、IP Configuration等)以及核心系統內的資料結構(例如：路由表)的改變。而且也未連結核心函式的執行與網路封包的關連性，無法追蹤核心行為與網路協定的互動與整合效果，因此並不適用於分析網路通訊裝置的通訊行為。 我們在本年度執行中的計畫中，已經研發核心行為(核心函式與核心事件)之追蹤與量測技術，可以準確記錄核心函式執行與核心事件發生的時間點。此外，我們也分析網路封包擷取工具資料儲存的結構，做為網路協定與核心行為整合的基礎。最後，我們建構了VoIP與Web Application的測試例，預計在本年度完成，以作為建構測試環境與評比基準化之基礎。 延續本年度執行中的計畫的研發成果，本子計劃規劃了三個研究主題: (1)核心行為之追蹤與量測技術之研發；(2)核心行為與網路協定整合分析工具之研發及(3) 通訊行為整合測試與評比之基準化。預計以兩年的時程發展一套適用於嵌入式網路通訊裝置的整合分析工具。 第一個主題是要研發核心行為(包含核心函式執行序列與核心事件)與網路協定的整合追蹤與量測技術，以及將此追蹤與量測技術工具化。第二個主題是要研發藉由追蹤與量測的結果合成(synthesize)通訊行為(包含核心函式序列、核心事件與網路封包)的技術，開發通訊行為整合分析工具，並且提供不同解析程度的分析結果。第三個主題是要制定通訊行為的標準測試環境以及評比基準，讓通訊裝置的測試結果具有參考公信力。此外，在整個計畫執行過程我們會遵循開放原始碼的規範，落實開放原始碼化的目標。 計劃成果將包含核心行為的追蹤與量測工具、通訊行為的整合分析工具與基準化測試例與評比項目的制定，為嵌入式裝置開發者提供一套網路通訊系統的量測、分析與評比的工具與發展平台。

This two-year subproject aims to design and develop tools for analyzing and benchmarking kernel and protocols of embedded networking and communication devices. With these tools, developers of embedded devices can analyze the efficiency and discover the bottlenecks of the networking protocols and kernel software. While evaluating the communication system of embedded networking and communication devices, we need to consider not only the network accessibility but also the critical performance metrics, such as transmission delays, response times, and packet losses, of embedded networking devices. These metrics depend on the integrated behavior of both the networking protocols and kernel software of the devices. However, no tools so far can be used to analyze such integrated behavior on an embedded networking device. As we know that current network sniffers can be used to analyze the “external” networking behaviors, namely sequences of protocol packets exchanged, but not the “internal” behaviors of kernel software of the networking protocol stack. Furthermore, current kernel probing tools can track only the execution sequences and times of kernel processes, and contents of memory locations specified by users, but not the kernel events or signals and the changes in internal kernel data structures (such as IP table). In addition, these tools cannot be used to deduce the integrated effects of the internal behavior of kernel software and the external behavior of networking protocols. Currently, we have developed the tracking and measurement techniques for kernel behaviors (kernel procedures and kernel events.) In addition, we have also analyzed the profiling structures of network sniffers. Therefore, we are ready to investigate the integrated effects of kernel software and networking protocols. Finally, we are constructing two test cases for two represented applications, VoIP and Web services. The test cases will be finished by the end of this academic year, and will serve as the basis for this two-year subproject in designing standard benchmarking. In this two-year subproject, by leveraging the results of our previous research, we will focus on three research topics: (1) Design and implementation of tracking and measurement tools for kernel behavior, (2) Development of the analysis tools for the integrated effects of kernel software and networking protocols, and (3) Design of the standard benchmarking and metrics for the integrated effects of kernel software and networking protocols. During the course of the project, we will follow the regulation of CMM/CMMI.