Abstract: The load balancing problem of network resources in large-scale ERP systems is studied.Firstly，the demand of this issue is modeled and analyzed theoretically;secondly，a heuristic objective function which satisfies the host and network constraints is designed，and furthermore the model is transformed into the problem of degree-constrained minimum spanning tree;thirdly，a simulated annealing algorithm is designed to deal with the problem and a load balancing scheme based on the algorithm named LABS is proposed eventually.According to the theoretical analysis，the time complexity of the computation operated by the whole network is the square of the size of node.The simulation results show that，by selecting appropriate factors，the scheme not only can absorb most of the nodes to participate in load balancing operations，but also can reduce the bottlenecks of nodes and the average utilization rate of resources significantly.

Key words: Enterprise Resource Planning（ERP） system, simulated annealing algorithm, load balancing, distributed

摘要： 针对大型ERP系统的网络资源负载均衡问题展开研究。在理论上对问题的需求进行建模分析，设计了满足主机和网络性能约束的启发式目标函数，将模型转化为度约束最小生成树问题，设计了一种模拟退火算法对此问题进行处理，提出了基于该算法的资源负载均衡方案LABS。理论分析表明，整个网络执行该方案的时间复杂度为节点规模的平方阶，说明方案具有较强的可用性与可伸缩性。仿真实验结果显示，通过选择适当的启发因子，算法不仅可以吸纳大部分节点协同参与负载均衡操作，还能够显著减少系统中的瓶颈节点数，降低平均资源使用率。

关键词: 企业资源计划（ERP）系统, 模拟退火算法, 负载均衡, 分布式