一個顧及服務品質且基於門檻的雲資源配置框架

Abstract

對於提供雲端相關服務而言，資源配置是一個非常重要的議題。如果資源配置的管理不夠精準，雲端服務可能會面臨資源不足或是資源浪費等情形。自動延展(Auto Scaling)是一種常用於基礎設施即服務(IaaS)中的資源配置方法，它可以維護資源的數量，避免在雲端服務需求增加或遞減時造成資源浪費或不足的情形，因此有非常多的雲端服務支援自動延展的功能。然而，自動延展依舊有其缺點，其中之一是它無法預測使用者的使用感受。像是當運算單元減少的情況下，當使用者面臨到效能低落時所產生的負面觀感。 有別於以往將實體資源映射到虛擬資源的資源利用的優化方式，本論文將專注於軟體層的資源配置的最佳化，提出一個新穎的雲端資源管理框架。此系統除了可以動態地自動延展，還可以藉由偵測使用者端的回應時間，給予使用者較佳的使用感受。此系統的目標便是以雲端軟體負載量的變化量為基礎，優化虛擬資源的配置。為此，我們定義了許多種不同條件的門檻，像是CPU使用率、回應時間及錯誤率，以期藉由不同條件的組成，在顧及服品質的前提下，動態地調整雲端資源的配置總額。在論文的最後，我們將藉由實驗的方式，證明我們所提出的系統能確實達到資源利用的優化。

Resource allocation, which is the process of assigning the amount of resources needed by cloud applications, is very crucial concern in the service provision. If the allocation of resources is not managed precisely, the cloud services may starve during the peak load time or waste the resources during the off-peak time. Auto scaling mechanism is one approach used in IaaS in which service providers can maintain the resources and reduce waste resources by automatically increase or decrease them when needed. Some of the cloud services support auto-scaling functionality. However, it is still difficult to predict the client side experience which later will cause in decreasing performance because of lacking computing instances. In this paper, we focus on allocating resources at the application level rather than mapping the physical resources to virtual resources for the efficient resource utilization. We present a novel cloud resource management framework which supports dynamic auto-scaling. The proposed system will monitor the end-user’s response time directly from client-side. We defined several thresholds with Quality of Services (QoS) considerations which include response time and error rates sampling in order to optimize the decision of reallocating the virtual resources. The goal of this study was to dynamically allocate the virtual resource among the cloud applications based on their workload changes in order to improve resource utilization, prevent from resource starvation and reduce the user usage cost. The experimental results show the proposed framework can improve the efficiency of resource utilization.