Research Of Service Based Resource Allocation In Networks

With the development of network, there are many new challenges in the research of network resource allocation, such as: the scarcity of resources, the diversity of services, the distribution of resources and the commercialization of applications. Nowadays, network providers pay more attention to the satisfaction of customers; the most important thing for them is the services. Traditional researches of resource allocation attach importance to the schedule of packets, providing the QoS guarantee, and so on. It hasn't considered the effect of follow factors, such as: the payment ability of users and the satisfaction degree of users. Accordingly, the result of resources allocation is not the most adaptive and optimal. In order to solve these problems, the thesis introduces the price factor into the research of network resource allocation, and then brings out a method based on service price. The method also can provide QoS guarantees. The thesis also gives a mechanism to guide the users making proper choices of end-to-end service classes.Firstly, the thesis presents a Market and QoS based Network Resource Allocation Model, named as MQNRA Model. This Model introduces the price factor into network resource allocation, which inherits the advantages of Differentiated Services and Integrated Services Model. It reconstructs the Bandwidth Broker and adds control module related to price. By this way, users can make choice of services based on price. It also provides the operation platform for optimal allocation.Aim at single ISP region, the thesis provides a method to gain the optimal resource allocation with QoS guarantees and also gain the highest satisfaction degree of all users. The typical SRA (Simple Resource Allocation) algorithm is so complicated that the control becomes a considerable expense. Therefore, the paper puts forward a pre-estimation based algorithm, named PE_ESRA ( Pre-Estimation based ESRA algorithm), which reduces the complexity and avoids unnecessary resources configure by introducing the mechanism of pre-estimation before the iteratively equilibrium price computing. Another disadvantage for SRA algorithm is that it cannot avoid the instance that some traffic classes may be jammed abruptly, but other traffic classes have free resources. The thesis presents an algorithm, called PEB_ESRA (Pre-Estimation and Borrow based ESRA algorithm),which permits borrowing resources among traffic classes. To some extent, it solves the short-term unbalance of resources allocation.Multidimensional resources may always be used in network application. The scarcity degree of distinct resources is different and the effect on QoS is also different. To attain the optimization of the whole system, the paper studies how to unite allocate the multidimensional resources and brings forward an economical model and algorithm.In order to achieve the end-to-end QoS, the thesis introduces the price factor into the process of end-to-end resource negotiation. It brings forward a new application level end-to-end QoS negotiation protocol, named SPNP (Service Price based Negotiation Protocol) . SPNP provides a well-designed way for users to communicate with the network system. It has the ability for query and quotation, also provides the mechanism to implement the EESSPA algorithm.The process of end-to-end resource negotiation must be concerned with the problem of QoS partition. The thesis provides EESSPA (End to End Service Select and Pricing Algorithm) algorithm, and illustrates it by the optimization of end-to-end delay allocation. Our algorithm may find the optimized end-to-end delay allocation. The result can satisfy the end-to-end delay requirement and minimize the end-to-end cost of user. EESSPA algorithm can avoid the information explosion by distributed control. Its computation complexity is independent of the number of hops along the path.Finally, the thesis summarizes our work and points out the next research direction.