面向海量业务场景的网络智能流量调度算法研究

Research on network intelligent traffic scheduling algorithm undermassive traffic environment

为解决海量业务场景下传统路由策略计算成本高、容易导致网络拥塞等问题,以最小化最大链路利用率为目标,提出一种网络智能流量调度算法以实现业务流的高效传输。基于业务量矩阵等网络实时状态,采用启发式算法计算当前网络状态下最优的流量调度方案;采用反向传播(back propagation,BP)神经网络构建网络智能流量调度模型,并根据网络实时状态与流量调度方案的映射关系对网络智能流量调度模型进行离线训练。在网络运行阶段,采用网络智能流量调度模型实时对业务流进行调度。实验验证表明,所提算法相比于传统启发式路由策略,最大链路利用率、平均时延、最大丢包率等指标都有较大提升。

With the deep integration of the Internet and society,a large number of emerging services have been introduced into the network.To solve the problem of high calculation cost of traditional routing policies and network congestion under massive traffic environment,a network intelligent traffic scheduling algorithm is proposed to realize the efficient transmission of traffic flow to minimize the maximum link utilization.Firstly,according to the traffic matrix,heuristic algorithms are used to calculate the optimal traffic scheduling scheme based on current network status.Secondly,back propagation neural network is used to construct the network intelligent scheduling model,and the traffic scheduling model is trained offline according to the mapping relationship between network states and traffic scheduling scheme.In the network operation stage,the intelligent network traffic scheduling model is used to schedule traffic flows in real time.Experimental verification shows that,compared to traditional heuristic routing strategies,the proposed algorithm has significantly improved indicators such as maximum link utilization,average latency,and maximum packet loss.