Research On Key Techniques Of Network Resource Management

With rapid development of Internet and large-scale applications of computer networks in recent years, the demands for resource sharing and interactive cooperation have been increasing in open distributed processing environment. The resources of different institutions in Internet are being abstracted and shared with such techniques as network and cloud computing, so that Internet has become a global gridding and a resource-sharing platform in the distributed, large-scale and heterogeneous networks. As a result, intensive data transmission and large-scale computing applications come into being. Network and cloud computing apply virtual resource pool dynamically to coordinate network resources, so that users can use all kinds of resources distributed in Internet apparently, just like logging in a supercomputer. Thus, the effective management and resources sharing in network become an important study task in the large-scale distributed computing environment.In this background, the key techniques of resource sharing management in network are studied in the projects "Study of the XXXXXXXXX Network Sharing Technology, the Study of the XXXX Network Attack and In-depth Defense System and the Online Sharing Platform Development of Shenmu County Elementary Education Resources". This paper focuses on three questions of network resource sharing mechanism, they are security trust, task scheduling and resource server fault management.In this paper, the concept Trust is introduced to build a safe and dependable strategic model based on dual identity authentication, the purpose is guaranteeing resource sharing and security of both parties in distributed opening environment to support the applications of network resource sharing. The corresponding algorithm and its performance are analyzed and verified. In order to address the problem that the existing task-allocation algorithm is easy to fall into the locally optimal solution and has low rate of convergence, a dynamic task resource allocation algorithm is put forth in a combination of the greedy algorithm and ant colony algorithm. Because the server cluster is a basic guarantee of the network resource sharing, the fault early warning mechanism of the network resource server cluster system is proposed based on the event association rules, it is used to ensure a stable operation of server clustering and improve management of the whole server clusters. The feasibility and performance of the scheme are verified by building the test prototype.The main research work and results are as follows:(1) Construct a security strategic model based on trusted network resource so as to guarantee security of resource cooperation by security certification and dependable mechanism. Aiming at the security of user certification in the distributed computing, a security strategic model is put forth based on dual identity authentication, and the authentication discussion is done respectively for users in the different distributed environment. Based on properties of the one-way hash function, the access control policy is established for identifying authentication by the type of users, and a mutual authentication relation is set up between users and servers. The security analysis show that the authentication model proposed in this paper not only ensures legal identity of both parties but also interdicts access of illegal users to the resource data. By this way, user safety and data privacy are well protected.(2) Build a dynamic access control model of trust management mechanism in the autonomous domain to prevent user from deceiving and stopping illegal resources. The trust management is performed by the autonomous domain as a unit, and this mechanism is defined in a corresponding trust. The computation of trust relation between domains and in the intra-domain, as well as trust evaluation, is given. The analysis results show that the trust management mechanism in the autonomous domain put forth in this paper can address the safe access to resources effectively in distributed network environment. As a result, the access control model in the gridding environment has the higher expansibility and feasibility, so that the gridding resources can be used more rationally and effectively.(3) Propose the cloud computing task scheduling model integrating the greedy algorithm with ant colony algorithm, and offer the mathematical descriptions of the model. The solution of this model can overcome the lack of initial pheromone of ant colony algorithm, but also to take advantage of the positive feedback mechanism of ant colony algorithm to accelerate the evolutionary process; so that final solution tends to the global optimum. The original solution is generated quickly by the fast convergence of the greedy algorithm to update initial pheromone quantity of the primary and secondary nodes of the ant colony algorithm, so that the solution in a combination of two algorithms tends to be a globally optimal one. The full node distance is taken as a greedy wandering condition, and the node in a smallest average distance in topological graph is obtained as the a main job scheduling node according to the thought of optimal solution found by each step of the greedy algorithm, then resource allocation is done for the user job. With the autonomous domain introduced, the appropriate distributed domain and domain job scheduling nodes are obtained by the clustering algorithm. The initial pheromone of the ant colony algorithm is updated based on main job scheduling node and domain job scheduling nodes in autonomous domain. The heuristic information of ant colony algorithm, ant route elective probability and pheromone updating rules are given.(4) Put forward the initial rule of node pheromone and the re-initial rule of local pheromone. The initial value of ant task scheduling algorithm is generated quickly by the fast convergence of the greedy algorithm. And the initialization rules are offered. When the failure nodes or the newly-added nodes appear, in order to find fast the optimal route in the network with the failure node or the newly-added node, the re-initial rule of local pheromone is proposed. With a re-initialization of local pheromone nearby the failure node or the newly-added node, most original optimal route information can be maintained, and the new route can be found locally. Finally, the algorithm is verified by test.(5) Set forth the fault early warning mechanism of network resource server cluster system based on event association rules, which can position early failure source quickly and eliminate the failure hidden dangers timely before the failure occurs, so as to ensure stable and continuous services of resource servers. The object-oriented method is applied to structure the quintuple to store the failure event and get failure decision algorithm. The association rules of early warning event are obtained based on the SCFEWS_Apriori association rules mining algorithm and its relevance tree is built by the algorithm. The failure filter model algorithm based on early warning event relevance tree is proposed to filter redundant failure event and reserve the failure source event as much as possible. It has been verified through performance test that the early warning mechanism put forth in this chapter can improve accuracy of failure positioning effectively.