This paper proposes a scheme to practically resolve the load imbalance state for orchestrated multiple controllers architecture in software-defined networks (SDN). Multiple controllers architecture is crucial to SDN implementation in large scale networks because this architecture provides efficient performance for end-to-end services, such as reliability and scalability in SDN networks. However, when there are multiple SDN controllers in the network, a load imbalance state may occur. The load imbalance problem can notably degrade service level in some parts of the network because the SDN controllers in those network areas have much higher amount of processing load. Existing works solved the load imbalance problem by migrating the load to multiple SDN controllers in order to maintain an acceptable level of load in all SDN controllers. Nevertheless, most of these works did not consider propagation delay and processing time in their load definition. In large-scale networks, high propagation delay is likely to cause late response from the SDN controller, which may result in a degraded performance in the SDN networks. In this paper, a new load balancing scheme is proposed. The proposed scheme is formulated as an integer linear programming problem (ILP). It defines SDN controller’s load based on propagation delay, processing time at the controller and the number of request messages in order to provide an accurate representation of load in practical environments. Generally, ILP may take a long period of time to process. Therefore, a heuristic algorithm that bases on the proposed load balancing scheme is also developed to provide shorter processing time. Computer simulations and practical implementation in Pica8 switch show that the proposed scheme reduces the average maximum load by at least 9.85%, compared to a conventional scheme.

本文提出了一种方案，以实际解决软件定义网络（SDN）中编排多控制器架构的负载不平衡状态。多控制器架构对于大规模网络中的 SDN 实施至关重要，因为这种架构为端到端服务提供了高效的性能，例如 SDN 网络中的可靠性和可扩展性。但是，当网络中存在多个SDN控制器时，可能会出现负载不平衡状态。负载不平衡问题会显着降低网络某些部分的服务水平，因为这些网络区域中的 SDN 控制器具有更高的处理负载量。现有工作通过将负载迁移到多个 SDN 控制器来解决负载不平衡问题，以便在所有 SDN 控制器中保持可接受的负载水平。尽管如此，这些作品中的大多数在其负载定义中都没有考虑传播延迟和处理时间。在大规模网络中，高传播延迟可能会导致 SDN 控制器的延迟响应，这可能导致 SDN 网络的性能下降。本文提出了一种新的负载均衡方案。所提出的方案被表述为整数线性规划问题（ILP）。它根据传播延迟、控制器的处理时间和请求消息的数量来定义 SDN 控制器的负载，以便在实际环境中提供负载的准确表示。通常，ILP 可能需要很长时间来处理。因此，还开发了基于所提出的负载平衡方案的启发式算法以提供更短的处理时间。