軟體定義網路的概念，從 OpenFlow 問世後，更為人所知。在過去軟體定義網路就常被應用在虛擬化網路，OpenFlow 的興起，讓軟體定義網路除了虛擬化網路外，連實體的網路設備都能夠進行管理，透過分離控制層與網路層，使網路的效率能夠更加快速更佳流暢。在近幾年，許多研究都以 OpenFlow 進行研究方向，都在探討 OpenFlow 是否能取代現今的一般網路，以及探討是否能透過 OpenFlow 對網路進行一些加值的服務，例如負載平衡、防火牆等。負載平衡的加值服務一些製作 OpenFlow 控制器的提供者已經有相關的套件可以使用，在過去虛擬化網路的時候就已經有相關的服務存在，如 OpenStack 的 Neutron，但他們的機制還是類似舊有的負載平衡機制，在每次新的連線進來後都會去詢問控制器，而造成額外的延遲。因此本論文使用遮罩的方式，直接在交換器或路由器上就直接進行負載平衡的機制，並加入使用者預判演算法，動態的更改遮罩範圍，讓這個機制可以更加符合各項服務的實際情形。在我們的實驗中，測試了不同預測演算法對於流量的預測情形，導入真實的服務流量來測試我們的整個負載平衡模組，並在最後與其他的負載平衡模組進行延遲以及平衡情形的比較。雲端技術的應用和服務與日俱增，因此在本論文之中，將所提出來的機制實做在雲端服務上，證明我們的機制是可以使用在各種服務平台上。 With the advent of OpenFlow, the concept of Software-Defined Network (SDN) becomes much popular. In the past, SDN had been often used for network virtualization; however, with the rise of OpenFlow, which speeds up network performance by separating the control layer from the data layer, SDN can be further used to manage physical network facilities. In recent years, many researches focus on using OpenFlow to substitute normal networks and look for adding value-added services with OpenFlow, such as the load balancer or the firewall. Currently some OpenFlow controller providers have already provided users with load balancer packages in their controllers for virtual networks, such as the Neutron package in OpenStack; nevertheless, the existing load balancer packages work in the old fashion that causes extra delay since they poll controllers for every new coming connection. In this thesis, we use the wildcard mask to implement the load balance method directly on switches or routers and add a user prediction mechanism to dynamically change the range of the wildcard mask; in this way, the load balance mechanism can be applied conforming to real service situations. In our experiment, we test the accuracies of flow prediction for different prediction algorithms and compare the delay times and balance situations of the proposed method with other load balancers.With the popularity of cloud computing, the demand of cloud infrastructure also increases. As a result, we also apply our load balance mechanism on cloud services and prove that the proposed method can be applied on varieties of service platforms.
