In this paper, we present a traffic-aware flow migration approach, which reduces data plane load in Software-Defined Networking (SDN) during a network update. SDN update involves rerouting of multiple traffic flows to accommodate new flows. An unplanned flow migration schedule overloads the data plane by burdening the data links and flooding the rule-space of capacity-constrained SDN switches. The overload of data links and switches blocks the update process, and the network fails to address the Quality of Service (QoS) demands of the traffic flows, especially latency-sensitive flows. Prior approaches migrate flows without considering load reduction of the data plane along with QoS demands of the flows. In this work, we propose a load reduction strategy that prioritizes traffic flows based on QoS demands and aims to avoid link congestion and rule-space overflow during flow migration. The proposed scheme significantly reduces the maximum data link bandwidth usage. In particular, the maximum data link bandwidth usage is 13.22% less than the two-phase update approach.

中文翻译：

---

DART：减少 SDN 中流量迁移的数据平面负载

在本文中，我们提出了一种流量感知流迁移方法，可在网络更新期间减少软件定义网络 (SDN) 中的数据平面负载。SDN 更新涉及重新路由多个流量流以适应新流量。计划外的流迁移计划会加重数据链路的负担并淹没容量受限的 SDN 交换机的规则空间，从而使数据平面过载。数据链路和交换机的过载会阻止更新过程，并且网络无法满足流量流的服务质量 (QoS) 需求，尤其是对延迟敏感的流。先前的方法迁移流而不考虑数据平面的负载减少以及流的QoS需求。在这项工作中，我们提出了一种负载减少策略，该策略根据 QoS 需求对流量进行优先级排序，旨在避免流量迁移期间的链路拥塞和规则空间溢出。所提出的方案显着降低了最大数据链路带宽使用。特别是，最大数据链路带宽使用比两阶段更新方法少 13.22%。