Key Techniques Study On Resource Management On Dual-mode Terminals In Cable Television Networks

With the development of digital television service in our country, dual-mode terminals(dual-mode set-top boxes) are becoming widely used digital terminal devices in people’s homes. With the development of bidirectional transmission systems on cable networks, HFC access networks are becoming the popular access networks for dual-mode terminals. Bandwidth resource in HFC access networks is becoming a critical factor of the Quality of Service of applications on dual-mode terminals. The convergence of three networks, makes a lot of converged services running on dual-mode terminals, and makes dual-mode terminals develop to smart television sets with open software architectures. Concurrent application tasks on smart television sets will compete for system resources (such as CPU, memory, tuners, etc.) and network bandwidth. Therefore, more efficient methods of resource management are necessary on dual-mode terminals, for guaranteed Quality of Service of applications. At the same time, it is necessary to appropriately allocate bandwidth of HFC access networks, so that the bandwidth requirement of applications on dual-mode terminals will be satisfied.Under the support of the National Key Technology R&D Program "new generation of radio and television service system supporting cross-regional and multi-service-providers" and National863project "The development of service architecture of convergent networks", this dissertation will discuss serveral key issues ralated to resource management on dual-mode terminals and bandwidth allocation in HFC access networks. The main contributions and innovations are as follows:1) In the past, resource allocation model in real-time systems are usually based on discrete resource-utility options. In some other researches, QoS or utility of applications are described as continuous functions about system resources. However, pure discrete resource-utility mappings or pure continuous resource-utility mappings, are not ideal models for describing the resource-utility relations on dual-mode terminals. We proposed a notion named "continuous utility subset", which not only describes some dicrete resource-utility mappings, but also describes some continuous resource-utility mappings, so it is an ideal model for dual-mode terminals. Based on the notion of continuous utility subsets, we make two innovations in the resource allocation algorithms on real-time systems, which are presented as follows:(i) In real-time systems, utility of an application will gradually become saturated when a certain amount of resources are allocated to it, so the resource consumption functions will gradually become concave functions. This character is usually related to the discrete resource-utility options in a same "continuous utility subset". According to this character, based on an existing heuristic algorithm HEU, we proposed a new heuristic algorithm called T-HEU, which can reduce the time complexity of HEU. The time complexity of HEU is O(MN2L2), and the time complexity of T-HEU is between O(MNL) and O(MN2L2).(ii) The resource allocation algorithms based on discrete resource-utility options leave some residual resources unused. This causes waste of some resources(such as CPU, network bandwidth, etc.). By defining continuous utility subsets, the allocation of residual resources can be modeled as linear programming problems. A heuristic algorithm RRA_HEU is proposed to achieve near-optimal solutions. Simulations show that when the number of tasks is smaller than15, RRA_HEU executes in a shorter time than simplex method and primal-dual interior point method. When allocating residual resources to a large number of tasks, simplex method is an appropriate algorithm, which executes in shorter time than primal-dual interior point method. By applying the residual resource allocation algorithm to the resource allocation for application processes on dual-mode terminals, the use ratio of resources is increased, so the system utility is increased.2) We propose a new algorithm for UMM(Utility Max-Min) fair bandwidth allocation on a single link, called AUMMAA. This algorithm is more accurate than the UMM allocation algorithm based on piecewise linear utility functions, and the time complexity of them are similar.3) Based on AUMMAA algorithm, a QAM resource allocation scheme is proposed for utility fairness among application flows. QAM channel selecting algorithm QCSA is proposed to select a QAM channel for each elastic flow, and AUMMAA is applied to allocate the bandwidth on a single channel. Theory and simulation show that QCSA tends to increase bandwidth use ratio when the channel group has abundant available bandwidth, but emphasizes fairness among different channels when the channel group is heavily loaded.4) Base on QCSA algorithm, an bandwidth allocation scheme is proposed for fairness among different terminals. Weighted max-min fair allocation algorithm is applied to allocate bandwidth to different terminals. When allocating a terminal’s bandwidth to its own online applications, we proposed NOTF algorithm, which achieves a near-optimal aggregate utility of these applications, while promoting fairness among these applications. Simulations show that NOTF reduce the variance of utility of different applications by50%, while reducing the aggregate utility of them by only5%.