Plug-in Packet Scheduling Algorithms. Multi-service Network

The existing packet scheduling algorithms are mainly based on packet-by-packet method, in which packet transmission process can not be interrupted. When dealing with variable length packets communication, if a packet of RT (real-time) services arrives after the beginning of a NRT (non-real time) packet's transmission, traditional packet-by-packet scheduling may block the packet of real-time service in the packet queue until the transmission of the long non-real-time packet is finished and therefore has a bad effect on the delay and delay variation of the real-time service. Moreover, the higher the proportion of NRT services bandwidth ratio is, the worse RT services delay and jitter performance will be.In order to address the above problems, provide better QoS guarantees, and reduce the impact of NRT load on the performance of RT services, an inserting model packet scheduling algorithm is proposed in this paper. The key of this algorithm implementation is creating and analyzing packet insertion, restoring control messages, and detecting the beginning and end time of inserted RT packets. This paper chose the control code of 8B/10B coding in 1000 BASE-X PCS sub layer of Gigabit Ethernet as packet inserting and restoring control messages. And for detecting the beginning and end time of inserted RT services, totally three inserting model packet scheduling methods, which have different advantages and disadvantages, are proposed in this paper.The first is priority-based inserting model packet scheduling algorithms. In priority-based inserting model packet scheduling algorithms, low-priority transmission will not affect the high-priority transmission at all. High priority packets can be sent out as long as it is at the head of the high-priority queue, without waiting for the current low-Priority transmission to be finished. We selected some suitable simulation models and built those models on OPNET simulation platforms to evaluate the performance of this kind of inserting model packet scheduling. The simulation results quantitatively demonstrate the feasibility and effectiveness of this algorithm, but also show some shortcomings of it in the fairness and independence. The second is the scheduling list based inserting model packet scheduling algorithms. The inserting model packet scheduling algorithms based on the scheduling list shapes and schedules the service flows with variable length packet by using simple scheduling list. According to different QoS requirements and services characteristics of different services, scheduling list based inserting model packet scheduling algorithms employs different packets sending methods. This kind of algorithm can guarantee different delay and delay jitter requirements of RT services.The third is ShVC-based inserting model packet scheduling algorithm. After applying ShVC packet scheduling algorithm, the ShVC will determine the order and length of sending chance for RT and NRT services. While the ShVC-based inserting model packet scheduling algorithm can reduce complexity and overhead of ShVC, and in the same time make the RT transmission service not be affected by the NRT any more.According to the scheduling list and ShVC inserting model based packet scheduling algorithm, we built two different OPNET simulation models and used the same services simulation models to test them. The statistic results of OPNET simulation on the feasibility and effectiveness of these algorithms show that inserting model packet scheduling algorithm can significantly reduce the delay and delay jitter of high-priority packet, while will not reduce the bandwidth utilization, and therefore is strongly practical.