摘要
雲端資源的租用最主要的計價方式有分為隨取所需與資源保留兩種。隨取所需的資源通常以使用虛擬機器的時間每小時計價，用多少算多少非常有彈性；而資源保留的計價方式需付一次性的長期保留費用，保留的資源如有真正啟動需再支付虛擬機器使用的價格但較隨取所需的租用方式便宜許多。網頁應用服務的提供者利用雲端資源提供服務，需考慮資源使用的成本與滿足和使用者訂定的服務層級協議(Service Level Agreement; SLA)。因此如何決定一個雲端資源租用的策略非常重要。我們提供了兩階段的解決方案，給網頁應用服務提供者一個租用雲端資源的參考。第一階段為長期資源租用策略，我們推導出雲端資源租用成本的模型，透過此模型計算最佳的資源保留量。第二階段為動態資源啟用策略，我們利用Hidden Markov Model (HMM)為我們的資源使用量預測模型，預測資源需求，提早取得資源以克服雲端虛擬資源自發出要求到取得的延遲，滿足與使用者的SLA。我們透過真實的流量驗證我們的方法。研究結果顯示，我們的方法可有效的降低資源租用的成本。
並列摘要
Two kinds of pricing Model of cloud resource subscription have been proposed, namely on-demand pricing and reserved pricing. On-demand pricing charges subscriber per hour consumed for each launched virtual machine (VM). Reserved pricing charges a low, one-time payment for each VM reserved, and then a discounted price per hour usage for each VM. To provide Web application service by using cloud resource, a service provider needs to consider the resource cost and Service Level Agreement (SLA) between the provider and its users. Therefore, it is important to plan a suitable cloud resource subscription policy. We propose a two-phase solution to solve the cloud resource subscription problem. The first phase considers long-term resource reservation. In this phase, we obtain the optimal long-term resource reservation policy based on a real-world cost model. The second phase is dynamic resource subscription phase. In order to overcome dynamic resource demand, in this phase, we use Hidden Markov Model (HMM) to predict resource demand which is then used to allocate VM resource adaptively. We evaluate our solution using real-world resource demand data. Our numerical results indicate that we are able to reduce the cost of cloud resource subscription significantly.