可抽換式元件架構之雲端作業系統

Abstract

雲端運算系統透過虛擬基礎設施資源來提供基礎架構層的服務給用戶使用，這些基礎設施資源包括計算單元、記憶體、儲存系統和網路系統。每個類別的基礎設施資源都是雲端運算系統中的一個子系統，而雲端運算系統則調配這些基礎設施子系統，來為用戶提供服務。 目前大多數雲端運算系統的基礎設施子系統以及決策演算法元件皆缺乏抽換性的設計。缺乏元件可抽換性將限制基礎設施子系統以及決策演算法的發展。我們認為，雲端運算系統應具有靈活性，任由內部基礎設施子系統以及決策演算法自由的更換。 本文介紹了大王椰子，是一個階層式架構雲端運算系統，並具有元件可抽換式設計。大王椰子基於各種特別設計的元件介面，來達到子系統的可抽換性。元件可抽換的能力使管理員能夠靈活地選擇最適當的基礎設施子系統，同時可以促進雲端運算系統各方面的發展。 大王椰子亦提供雲端運算系統決策演算法的實驗平台。這些演算法完全與大王椰子內部所有的子系統獨立，又可以輕易的套用進大王椰子來進行實驗。因此設計者可以專注於演算法的設計，無需擔心如何與大王椰子整合。我們相信，元件可抽換性在雲端運算系統中是一個非常重要的議題。

Cloud computing system provides services and manages infrastructures in data centers. The common infrastructures subsystems in a data center including: hypervisors, storage systems, network systems... etc. These infrastructure subsystems cooperate together to provide infrastructure level services. To provide services, cloud computing system needs to coordinate of the works between subsystems. On the other hand, current cloud computing systems are restricted in not only their favor infrastructure subsystem, but also decision algorithms. This limitation restricts the development of decision algorithms. This limitation also restricts infrastructure subsystems to the one embedded in current cloud computing system. The cloud computing system should have flexibility to easily switch between different infrastructure subsystems and between different decision algorithms. Roystonea is a hierarchical distributed cloud computing system with pluggable component architecture. The component pluggable ability of Roystonea relies on the specific interfaces between Roystonea controlling system and infrastructure subsystem components. Roystonea controls subsystems by interfaces. It is the bridge of Roystonea controlling system and user specified components. Interfaces give administrators the flexibility to switch appropriate subsystem on their own. The interfaces also encourages development of infrastructure subsystems in cloud computing. With the flexibility of pluggable component, Roystonea also provide a test bed for decision algorithms, such as virtual machine provisioning, power conservation, and image file management, in cloud computing. Any language of decision algorithms can easily applied for Roystonea. We believed that the pluggable component is one of the most important concept of development in cloud computing system.