在MPLS網路流量工程上具有優先權的動態負載平衡演算法

Abstract

多重協定標籤交換技術(Multi-Protocol Label Switching，MPLS)除了保留傳統IP網路很好的連接性（connectivity），且加速了IP封包在路由器中的遞送(forwarding)，尤其它具有很強的流量工程(traffic engineering)能力，可提高網路資源利用率(network resource utilization)以及服務品質(QoS)。本論文的主要研究重點在於利用MPLS之限制性路由標籤發送協定(Constraint-Based Routed Label Distribution Protocol，CR-LDP)技術，找尋一種演算法，可應用在資料流具有優先權(priority)的情況下，根據資料流的優先權，給予合理、公平的資源分配，並使網路資源有效的被運用。 在我們提出之具有優先權的動態負載平衡演算法（Dynamic Load Balance Algorithm with Priority，DLBP）機制裡，當入口標籤交換路由器（Ingress LSR）收到一個新的資料流(traffic flow)時，我們會替它在MPLS網路裡建立一條路徑最短的而且可用頻寬符合需求的標籤交換路徑（LSP），但當網路頻寬不夠時，會盡量重新路由(reroute)其它資料流至合適的路徑，而騰出足夠的可用頻寬給新資料流使用，但若無法利用重新路由其它資料流，而騰出足夠的可用頻寬給新資料流使用時，則可侵佔(preempt)優先權比新資料流低的資料流保留頻寬，但在考慮侵佔對象時，也要使網路資源使用及傳輸品質最佳化。由模擬結果證明DLBP演算法確實可使網路資源有效的被運用，並且根據資料流的優先權，給予合理、公平的資源分配。

MPLS (Multi-Protocol Label Switching) technique not only reserves the high connectivity of traditional IP network, but speeds up IP packet forwarding in a router. Especially, its powerful capability of traffic engineering improves the network resource utilization and Quality of Service. The main key point of this thesis is to use the CR-LDP (Constraint-Based Routed Label Distribution Protocol) technique of MPLS to search for an algorithm, which can fairly distribute resource for a traffic flow according to its priority ,and makes use of the network resource more efficiently . In the DLBP (Dynamic Load Balance Algorithm with Priority) mechanism we propose, when Ingress LSR receives a new traffic flow, it will select a route whose length is the shortest and Free-capacity is larger than or equal to the bandwidth request of the new traffic flow ,and then along the selected route to create an LSP in the MPLS network for the traffic flow . If the Ingress LSR fails to find such a route ,in our mechanism ,it will do its best to reroute some existing smaller traffic flows to another appropriate routes so as to reserve a high capacity route for the new traffic flow. Once the Ingress LSR still fails to find a sufficient bandwidth for the new traffic flow ,we allow the new traffic flow to preempt the reserved bandwidth of another one with lower priority. When selecting which one will be preempted, we need to consider how it can make the use of network resource and the transportation quality optimized. Our results of simulation prove that DLBP indeed makes network resource used effectively and the transportation quality optimized.