運用最佳化技術於異質雲端運算環境能源效率與資源整合管理研究

Abstract

現代資訊科技的進步造就了數位化和人工智慧的蓬勃發展，同時也推動著產業的轉型。然而，這種快速進步伴隨著對於雲端運算的需求，其大量的能源消耗也將對環境的影響日漸加劇，如何有效率的運用運算資源儼然成為一個重要的議題。本研究以綠色資訊科技(Green IT)中的運算能源消耗議題的角度切入，討論在各種雲端運算的情境下，如何透過最佳化演算法的分配有效運用其運算資源來達到節省能耗及加快工作效率之目的。 本研究將分為三個主軸討論，由具備異質運算資源的雲端運算中心出發，考量不同運算資源所適合執行之任務類型，將其在所需應用執行的時限內最佳化分配來降低能源消耗。在單一運算中心內部分配後，本研究將所討論之情境延伸至不同運算單元間的分散式雲端運算，考量在邊緣運算的環境下之運算、傳輸與等待問題，從運算端點到核心運算中心間如何透過運算任務之分配有效降低其能源消耗。此外，在環境監測和物聯網等無線感測網路應用時，由於使用電池供電的感測器和無線通訊存在其獨特的能耗挑戰，本研究也探索在樹狀感測的網路架構中最佳化能源消耗之議題。研究使用拉格朗日鬆弛法來開發在雲端運算各種情境中之最佳化演算法。期望在提高運算效率的同時減低科技發展對於環境的影響及營運成本，也提供未來資訊技術研究可以參考之觀點。

The contemporary landscape witnesses an unprecedented fusion of digitalization and artificial intelligence, driving transformative advancements across industries. However, this rapid progress is accompanied by a critical concern—the escalating environmental impact resulting from the intricate interplay between information technology and emergent industries. In response, Green Information Technology (Green IT) emerges as a strategic paradigm, prioritizing objectives such as mitigating environmental degradation and conserving energy resources. This research delves into efforts and strategies aimed at optimizing energy consumption, particularly within cloud computing environments. Motivated by significant energy consumption challenges associated with cloud computing and data transmission, exacerbated by growing data volumes and mobile device proliferation, this research focuses on enhancing energy efficiency and optimizing resource allocation within heterogeneous cloud computing centers. Aims to reduce total power consumption in distributed systems and improve energy management strategies for wireless sensor networks. Cloud computing, integral to modern infrastructure, offers scalability and cost-effectiveness but also drives substantial energy consumption within data centers. Additionally, data transmission requires significant energy, especially across geographical locations. Wireless sensor networks, crucial for applications like environmental monitoring and IoT, pose unique energy challenges due to battery-powered sensors and wireless communication. This research seeks to unravel energy optimization intricacies based on Lagrangian Relaxation within cloud computing across diverse scenarios not only mitigates environmental impact but also enhances operational efficiency and reduces costs in the digital age. The proposed optimization algorithms aim to address multifaceted challenges in heterogeneous landscapes, offering insights crucial for contemporary and future IT paradigms.