光纖網路(Optical Network)是最近幾年骨幹網路所採取的趨勢，具有傳輸速度快且不易受到電子訊號干擾等特性。光突發交換主要是把多個封包組合成一個burst，再利用光纖網路頻寬大的特質，一次把burst送出。由於光纖網路傳輸速度極快，故其各個交換機所需設定時間極短，獨立的控制信號技術(Out of Band)也應運而生，獨立控制信號技術的控制通道與資料通道是分開的，且控制信號必須先行送出然後接著再送出實際傳輸的資料，所以資料通道使用的排程方式便會影響到整個網路的傳輸速度，也是網路的瓶頸所在。本論文提出一個改進式的通道排程演算法，改善了最少可用未排程通道配合填空(Latest Available Unscheduled Channel with Void Filling，簡稱LAUC-VF)演算法，原有LAUC-VF演算法未考慮已經進入排程的burst，若burst太小而填空的空間大於burst甚多，則此填空會浪費許多空間，造成後來的burst無法有效利用。因此本論文所提之演算法考慮此填空大小是否最適於目前進入排程的burst，再決定其所排入之資料通道，以達到充分利用資料通道的空間，並符合資料通道排程的目的。本文中加入burst優先序的概念，使得優先序高的burst可以有較高的機會進入排程。另外，延遲線亦加入優先序的概念，優先序高的burst可以獲得較多的延遲線圈數，因此當通道擁塞時，其可進入排程的機率也會較高，且保證當網路壅塞時重要的資料仍可送出，以符合網路服務品質(Quality of Service，簡稱QoS)的要求。 Optical Network has become a trend on backbone network. It comes with great constitution including high transmission speed and not to be affected by electronic interference. The idea of Optical burst switching is to combine multiple packets into a single burst, making use of the wide bandwidth of optical network, and send out the burst at one sweep. The control channels and data channels are separated on Out of Band control signaling, and the real data are sent only after the control signal is on its way. This means the schedule method used over data channels would influence the overall transmission speed, and even become a bottleneck on the network. This paper proposes a improved channel scheduling algorithm to refine the Latest Available Unscheduled Channel with Void Filling, LAUC-VF. If the burst length is too small and the empty slot is too big for the burst, then the burst would become a waste for space, lowering the utilization for future use. The algorithm proposed in this paper sees into every channel with empty slots to find the best fit scheduling resolution. In this paper, we induce the idea of prioritized burst. Bursts come with higher priority have higher probability to be scheduled in-line. In addition, the fiber delay lines used here are also prioritized. Higher prioritized bursts can obtain more delay lines. While channels are congested, those bursts would be scheduled with higher probability, and be guaranteed to forward out. This keeps the QoS, working.
