Algorithmic Research Of Providing Guaranteed Services For Differentiated Service Network

Ⅰ. IntroductionInternet has experienced on explosively growth since 1990's. Widespread use of computers network, the information change from the single text data to the multimedia information such as the IP phone, distance teaching, distance medical treatment and so on, as well as the appearance of varied network application has made a significant problems of network congestion. So the single service can not satisfy the feature need of network transmission.In traditionally, IP network only provides the Best Effort service based on the TCP/IP protocol. So it doesn't satisfy the quality of the network for the application of the multimedia and the consumer. Such as the transmission delay, bandwidth and the data loss ratio, their performance can not be guarantee. So the Quality of Service (QoS) put forward. Differentiated Service (DiffServ) architecture is one kind of models presented by Internet Engineering Task Force (IETF) in order to provide QoS in the Internet. It provides different QoS on traditional IP network, especially the new services such as VOD, VOIP and video conference and so on, which need the relatively short delay, delay jittering, less loss ratio and wider bandwidth, must be realized through Internet network. Under the pushing of these applications how to arrange the IP network resource in the network and afford the QoS guarantee is becoming a hot topic at recent research in this field.Ⅱ. Improved Active Queue Management Algorithms in DiffServ (Ⅰ) RIO-CLCurrent assured forwarding (AF) service in DiffServ networks can provide stable guarantees in throughput, but is lacking of efficient schemes in ensuring queuing loss ratio. But nowadays, by increasing of real-time business such as multimedia traffic, it does not only request higher layer to correction coding management, but also request lower layers have certainly ability to permit the wrong. So we present a detailed study of an RIO-CL (RED with In/Out and Couple queue for Loss) algorithm which addresses both of these problems.The principle of RIO-CL is as follow: it adjust dynamic the maximum loss ratio threshold Tm iax of discard precedence i, it can guarantee RIO-C algorithm loss ratio performance on the dynamic or different network conditions. On the other side, configuring the router parameters in terms of target delay should be much easier than in terms of"abstract"parameters like queue thresholds, discard probabilities or averaging queue weights.The contrast of RIO-C and RIO-CL algorithm simulation experiment show: RIO-CL algorithm can assure the stability on the loss ratio, delay and average queue size. It can maintain the throughput, provide the stable guarantees loss ratio, and maintain the stable average queue size. This algorithm is applicable sensitive network in loss ratio.(Ⅱ) SARIO-CThe fundamentality problem of some algorithms is that they estimate the congestion only depending on the average queue size such as RED and RIO-C. Though the durative queue size implies the network congestion, it barely provides any information about the congestion degree. On the other hand, the calculation of the queue size has some fixed parameters. So we need a new algorithm, it can auto regulate the parameters along the network condition variety. So we bring forward the improve RIO-C algorithm―― assure the average delay and strengthen adaptive mechanism RIO-C algorithm: SARIO-C (Strengthen Adaptive RIO-C). this algorithm add the variation ratio besides the auto adaptive mechanism.The principle of SARIO-C is as follow: it adjusts the parameter maximal discard probability Pm ax according the variety ratio of the average queue size. But we design the adjust threshold for the sake of avoid adjusting of the natural queue fluctuate and bringing the queue heavy oscillations. When the variety ratio of the average queue size increase to some degree, we start adjust it. Certainly we also utilize the information of the average queue size. Because of make the average queue size approaches the half of buffer space and avoid Pm ax adjust too frequently; we define the target range of average queue size target. It is optimal keeping the average queue in the middle, and avoiding the queue empty or flood. The adaptation function of Pm ax uses the additive increase multiple decrease (AIMD).The contrast of RIO-C and SARIO-C algorithm simulation experiment show: the SARIO-C algorithm realizes the setting of the parameters, which follow the variation target automatically, at the same time, it maintains steady queue size and delay. especially the network condition variety is high frequency.Ⅲ. ConclusionAssured forwarding service in DiffServ networks can provide qualitative differentiation in ensuring queuing loss ratio and delay, but it is lacking of quantitative guarantee. In this paper, we put forward two different improvement algorithms, providing the quantitative guarantee for the queuing loss ratio and delay. It stress on the improvement of precedence reasonable expand and the setting of the parameters, the improvement of the parameter susceptive problem, and uncertain parameter infect the algorithm itself expansibility.The RIO-CL algorithm can maintain the throughput in RIO-C algorithm, and provide the stable guarantees loss ratio. This algorithm is applicable sensitive applications for loss ratio.The SARIO-C algorithm can guarantee the throughput and protecting the high precedence, and guarantee the stabilization of the average queue size and delay. In this way, when the network condition is variety, it can provide the stable delay and average queue size, especially the network condition variety is high frequency.