適用於低延遲服務且基於霧運算的多層資料處理及彙整架構

Abstract

隨著雲端運算在近代成為一項非常成熟的技術，人們的生活也越來越便利，因為雲端運算，人們不再被個人電腦的儲存空間和運算能力所拘束，但這個中央化的架構卻遇到了一些麻煩。我們都知道未來將會是物聯網的時代，這意味著生活中各式各樣的物品透過裝設感測器等等的方式都將具備連上網路的能力來智能化它們，而這些感測器基於成本上的考量可能都只具備一些簡單的功能像是只能上傳它們收集的資料，為了能讓它們能「智能化」勢必還要再搭配雲端運算去處理那些被上傳的資料，這些數量龐大的感測器和資料量將會是傳統的雲端運算架構的一大挑戰，由其是針對那些會與使用者互動、有即時性需求的應用。 因此，本碩論利用一種由雲端運算延伸出來的新架構-霧運算來解決剛才提到的雲端運算遇到的挑戰。事實上目前已經有相當多的論文將霧運算應用到各種不同的情境提出各自的架構，但大部分的論文都是以模擬的方式去測量他們的效能。在我們的論文中，我們利用霧運算實做了一個關於霧運算的原型，這個原型提供了一個霧運算架構應該具備的基本功能，任何想要使用這個原型去實做他們的應用的人只要根據他們的應用類型、需求加入特定的功能即可，我們也實際的將我們的系統放到真實的環境中去測量效能。

Recently years, there are a lot of interests in fog computing in the distributing computing field. Fog computing is a new computing architecture extended from cloud computing, and proposed to solve problems met on latency-sensitive and location-awareness IoT services. Although there are several fog computing-based theories and applications have been proposed, most of them only evaluated their works by theoretical simulations. Those are far from real situations and difficult to be applied into practice. Motivated by previous works, this study is aimed to propose a client-fog-cloud multilayer data processing and aggregation framework, based on fog computing paradigm. The proposed framework is designed to help latency-sensitive applications in IOT context, which meet requirements: widely distribution, massive uploading, low latency, and real-time interaction. Authors used the child abduction alert service as a sample to evaluate the proposed framework in practical scenarios, and compare performance and feasibility to the conventional cloud solution. Results showed that this framework can reduce about 32% response time and 30% data transferred to the cloud.