QoS Provisiong With Differentiated Queueing Service In Integrated Networks

With the mobile evolution of wireless broadband access networks and the broadband evo-lution of mobile cellular communication networks, the future networks are evolved to IP-basedintegrated networks with heterogeneous access technologies. Support of multiple access tech-nologies, provisioningofvariousapplicationsanywhereanytimeandanyhowthroughanythingwill been the important features of next generation networks. However, QoS provisioning in in-tegrated networks faces novel and serious challenges. For example, end-to-end QoS provision-ing becomes more complex due to various access technologies. The development of variousapplications requires QoS support of mixed flows and a wide range of delay guarantee. Thetime sensitive nature of the increasing real time applications makes delay bound violation prob-abilitybecomeanovelandimportantperformancemetricinmeasuringtheQoSprovisioningforintegrated networks. Furthermore, QoS over-provisioning is no longer suitable for integratednetworks due to the wireless resources constraint. Consequently, the study on the end-to-endQoS provisioning in such kind of networks is meaningful.Traditional QoS schemes include IntServ and DiffSev as well as their variants. However,besides that they are not able to balance well between scalability and QoS granularity, they alsoface a series of problems in QoS provisioning in integrated networks with the novel features ofintegrated networks mentioned above. Differentiated Queueing Service (DQS) was proposedto balance service granularity and scalability in wired networks initially and then extended towireless networks. The main service disciplines of DQS are as follows: DQS explicitly requireseach packet to carry its QoS requirements by itself in terms of end-to-end delay bound andpacket loss preference. Then the routers along the path properly place the arriving packets inthe queue according to their QoS requirements and those of all the already queued packets.Due to its per-packet service granularity, DQS could provide differentiated services accordingto end-to-end path situations and could support a wide range of QoS requirements especiallydelayrequirements, consequentlyitisasuitableQoSschemeforintegratednetworks. However,although DQS was suggested to be an end-to-end QoS scheme, more works on how to provisionQoSwithitinintegratednetworksshouldbedone,includingresourceallocation,QoS-constraintadmission control, how to support end-to-end QoS adaptability and so on. In the following we firstly overview the service disciplines of DQS, then briefly introducethe key technologies of DQS and the state of the art as well. In the sequel we will study theresource allocation of DQS for QoS provisioning and admission control for DQS with time-varying network capacity as well as end-to-end QoS performance models of DQS. Finally wepropose an adaptive end-to-end QoS provisioning scheme of DQS in integrated networks. Themain contributions are as follows:1) Input flows are modeled according to their delay requirements and a service capabilitycurve is proposed to represent the dynamic network capacity. Traditional methods on networkanalysis commonly model the Internet traffic in per-flow or per-class granularity. This papermodels the Internet traffic in packet granularity by delay bounds. Arrival traffic is classifiedinto limited data flows according to their delay requirements. Packets with the same delayrequirements are modeled into a traffic flow. This method can flexibly describe mixed QoSflow and support a wide range of delay requirements. We further propose a stochastic servicecapability curve to represent the dynamic wireless link with the center limit theorem. Thismethod can map the parameters of lower layer, such as channel capacity, SINR and mobilerate and so on, into service capability parameters of network layer including mean bandwidth,bandwidth variable and time interval. This method can simple the process of network calculusand reflect the property of time-varying bandwidth of wireless link.2) Studies on QoS provisioning of DQS with single node, including:a) Study on bandwidth and buffer allocation for QoS provisioning. We derive the re-quired bandwidth and buffer resources separately for the resource allocation manners ofper-flow dedicated resource allocation and global shared resources with multiple FBM(Fractional Brownian Motion) inputs and their delay requirements as well as constant net-work capacity. We also derive the required bandwidth and buffer resources for mixedtraffic, which include real-time and non-real-time applications.b) Study on admission control for QoS provisioning with time-varying network capac-ity. We build the QoS performance model and further propose a QoS guaranteed ad-mission control algorithm by using delay bound violation probability deduced above asa main threshold. Then SMDP (Semi-Markov Decision Process) are used to optimal theadmission control policy. 3) Studies on End-to-end QoS provisioning of DQS with multiple nodes, including:a) Analysisofend-to-endQoSperformanceofDQSwiththetheoryofnetworkcalculus.We firstly analyze the end-to-end QoS performance with deterministic service. A schemethat reserving end-to-end bandwidth according to envelope rate of input flow is proposedfor end-to-end QoS provisioning. Then an upper bound of end-to-end delay bound vio-lation probability is derived with stochastic network calculus and chernoff bounds. Theupper bound could be used to investigate the end-to-end QoS provisioning performanceand be applied to traffic control, thus it has important theoretical value.b) Proposed an adaptive end-to-end QoS provisioning scheme of DQS in integrated net-works. Weproposedistinctdynamicdelayboundestimationalgorithmsfordifferenttypesof nodes for QoS adaptation in integrated networks.The above proposals and models are all verification and validation by computer simula-tions. The simulation results show that all of them could achieve the desired results.