具高效率資源與服務管理之高可信度遊戲雲端計算---子計畫三：建構於雲端計算環境之大型多人線上遊戲資源配置機制

Abstract

計畫名稱：建構於雲端計算環境之大型多人線上遊戲資源配置機制 研究者：王國禎 經費來源：行政院國家科學委員會 由於大型多人線上遊戲(MMOG)技術日趨成熟且玩家人數也越來越多，因此每秒需 要處理的資料量及計算量也越來越多。基於此需求，MMOGs 非常適合建構在雲端計算 環境中。雲端遊戲牽涉到一個關鍵的研究問題是如何在雲端計算環境中，將資源做最有 效及公平的配置，即如何將這些龐大的資料量及計算量做出適當的伺服器配置，並且讓 伺服器間的負載量也能夠達到平衡。此外，資源配置的決策必需要快速判斷、公平且迅 速執行，才可以滿足玩家群的需求。這種高效率及公平的資源配置問題已經成為雲端遊 戲必須面對的重要研究課題。 在本子計畫中，我們提出一個適用於雲端遊戲環境之新的混合式網路架構模式，來 解決現今主從式網路架構中伺服器往往成為提升遊戲效能的瓶頸。首先，我們將伺服器 依不同功能作等級劃分。區域伺服器負責和玩家直接做連結，主要以主從式架構為主。 次級區域伺服器主要是選自性能較好的玩家設備，協助處理區域伺服器分派之不具風險 的資料與程式，以分攤區域伺服器的負擔。而區域伺服器之間的訊息傳遞主要是利用同 儕(P2P)方式，它是雲端計算環境的主要部分，把MMOGs 所需大量的運算與儲存資料 放入雲端。我們將以有效性(effectiveness)及公平性(fairness)來評估我們所提出的資源配 置機制。有效性的評估取決於上線玩家數目／回應時間的比值，而公平性指標(fairness index)將用以評估公平性。 本子計畫預計在三年內完成。在第一年中，我們將廣泛的進行相關文獻的探討，提 出新的資源配置方法，並模擬及評估其優缺點。在第二年中，我們將實作所提出研究 方法，並進一步改良及評估其合用性。然後，我們將與其他子計畫做初步整合與測試， 以作為資源配置方法之改進參考。在第三年中，我們將與其他子計畫做全面的整合與 測試，並以上線玩家數目／回應時間，以及公平指標來評估所提出之設計方法的整體 效能表現。本子計畫的預期研究成果為在支持大量上線玩家的前提下，達到快速回應 時間、玩家之間公平性及各區域伺服器負載平衡的研究目標，使伺服器不再是服務效 率的瓶頸。本研究計畫的執行成果，將可讓我們掌握雲端計算的關鍵技術，並幫助我 國之資通信及遊戲產業開創另一波新的契機。

Title: Resource Allocation for MMOGs in Cloud Computing Environments Principal Investigator: Kuochen Wang Sponsor: National Science Council Due to technology improvements of the massively multiplayer online game (MMOG), the number of players playing MMOGs is growing rapidly. The more active players play MMOGs on a game server, the more powerful the game server is required. Since the capability of a single server is limited, MMOG providers usually use a cluster of servers to deal with growing player demands. Existing architectures of MMOG server clusters may have player scalability and low load balancing problems; as a result, it makes cloud computing environments suitable for deploying MMOG server clusters. How to allocate resources in gaming cloud environments to achieve effectiveness and fairness is an essential research issue. The effectiveness involves high scalability and low response time. The fairness means gaming services received are irrelevant to the capabilities of gaming devices used. In this subproject, we propose a novel hybrid network model to resolve the server bottleneck problem of the client-server model traditionally used in gaming cloud environments. First, we classify gaming servers to several groups according to servers’ capabilities. Regional servers are directly connected to gaming players. When regional servers encounter overloading, secondary regional servers are introduced to help reducing the loading of regional servers. Secondary regional servers are selected from players’ gaming devices with better computing power. As to the communication between regional servers, P2P technology is used to achieve effectiveness (or scalability) and fairness. This subproject is expected to be completed within three years. In the first year, we will study the state of the art of MMOGs, cloud computing, and resource allocation. We will propose an efficient design approach of resource allocation and evaluate the strengths and weaknesses of our approach. In the second year, we will implement and evaluate the proposed approach and further improve the approach. Then, we will integrate our subproject with the other subprojects. In the third year, further integration and testing between our subject and the other subprojects will be carried out. Finally, we will evaluate the overall performance of our design approach in terms of the ratio of number of active gaming players to response time and the fairness index. Under the condition of having a large number of active gaming players, our research results can have fast response time, a high fairness index between active gaming players, and a high load balancing index among regional servers.