ATM网络带宽动态优化的广义粒子模型和算法  

A Generalized Particle Model and Algorithm for Dynamic Optimization of Bandwidth Allocation in ATM Networks

作　　者：帅典勋[1] 宫睿[2] 

机构地区：[1]华东理工大学计算机科学与工程系,上海200237 [2]清华大学智能技术与系统国家重点实验室,北京100080

出　　处：《计算机学报》2007年第3期380-396,共17页Chinese Journal of Computers

基　　金：国家自然科学基金项目(60473044;60575040;60073008);国家自然科学基金重点项目(60135010);清华大学智能技术与系统国家重点实验室的资助

摘　　要：提出一种新的广义粒子模型和算法,将ATM网络优化问题转变为对偶力场中粒子的运动学和动力学问题,从而分布并行地动态优化ATM网络的资源和带宽分配以及ATM网络的QoS通信合约.讨论了ATM网络动态优化的广义粒子模型的适应性、收敛性和稳定性等性质.ATM网络带宽分配优化问题是NP-完全问题.根据服务类型、通信流量特性和QoS参数,优化ATM网络的资源和带宽分配,对于提高网络吞吐能力、保证网络QoS性能有重要意义.理论分析和仿真实验表明,广义粒子模型和算法具有高度分布并行性,能体现资源需求的价格机制,能适应复杂的动态环境,易于硬件和软件实现.A novel generalized particle model （GPM） and its algorithm for dynamically optimizing both the VP bandwidth allocation and the VP negotiated QoS parameters in ATM networks are presented. The proposed approach transforms the dynamic bandwidth allocation problem among VP＇s in ATM networks into the kinematics and dynamics problem of particles in two reciprocal dual-force fields, so that the evolution of particles states can eventually results in an optimal solution of the original bandwidth allocation problem. The basic properties of GPM, including the suitability, convergency and stability, are discussed. The bandwidth allocation problem in ATM networks is NP-complete. Based on the service categories, traffic characteristics and QoS requirements, dynamically allocating the virtual path bandwidth in ATM networks plays a significant role in enhancing the ATM network throughput and improving the QoS performance. The theoretical analysis and numerous simulations on ATM network bandwidth allocation have shown that the GPM approach has the higher parallelism, lower computation complexities, easy of hardware implementation and better availability for complex environment.

关 键 词：ATM网络 带宽分配 广义粒子模型 分布并行算法 动力学过程 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]