摘要： 当前分布式系统中的负载均衡的预测都是基于对预测结果完全信任的机制，但是这种信任会有一定的风险，进而影响系统的效率。该文提出了一个基于对预测结果非完全信任的带有确认机制的，并且多种预测算法相结合、有一定的自适应能力的负载预测模型，并在该模型的基础上，结合了移动代理技术。模拟实验表明，该方法动态预测占用资源少、预测速度快、预测准确度也比较高，有很好的适应性，基本实现了动态负载均衡的目的。

关键词: 分布式系统；负载均衡；动态预测；自适应性

Abstract: Load balancing prediction in distributed system is based on full trust in prediction currently. However such trust may risk wrong prediction and degrade the systems efficiency. This paper proposes a no-full trust load prediction model with multiple algorithms and confirmation mechanism. Based on this model, it combines with mobile agent technology. The simulation testing shows that the dynamic prediction method produces faster and accurater prediction with less consumption of system resource and adapts well, which is a better method for load balancing

Key words: Distributed system; Load balancing; Dynamic prediction; Self-adaptation