Congestion Notification in Routed Lossless Ethernet Networks
End devices and their applications may not be aware of congestion in the network. A culprit device may continue to send (or solicit) more traffic on the network making the severity of congestion worse or increasing its duration. To solve this problem, the network switches can ‘explicitly’ notify the end devices as soon as they detect congestion. In response, the culprit devices can take preventive actions. 终端设备及其应用程序可能无法察觉网络拥塞。肇事设备可能会继续在网络上发送（或请求）更多流量，使拥塞的严重程度恶化或持续时间延长。为了解决这个问题，网络交换机可以在检测到拥塞时立即 "明确 "通知终端设备。作为回应，肇事设备可采取预防措施。
This approach of preventing congestion by notifying the end devices is available in all kinds of networks with different degrees of implementation and adoption. Chapter 6, the section on Congestion Prevention by Notifying the End Devices, provides details of implementation in Fibre Channel fabrics using Fabric Performance Impact Notifications (FPIN) and Congestion Signals. Chapter 8, the section on Congestion Notification in TCP Storage Networks explains this approach in TCP/IP storage networks. 这种通过通知终端设备来防止拥塞的方法适用于各种网络，其实施和采用程度各不相同。第 6 章 "通过通知终端设备防止拥塞 "一节详细介绍了在光纤通道 Fabric 中使用 Fabric Performance Impact Notifications (FPIN) 和 Congestion Signals 的实施情况。第 8 章 "TCP 存储网络中的拥塞通知 "一节介绍了 TCP/IP 存储网络中的这种方法。
As previously mentioned, the implementations and adoption of this approach in FCoE and RoCE environments are uncommon. Hence, this section does not explain them. The focus of this section is on congestion notification in routed lossless Ethernet networks for RoCEv2 Congestion Management (RCM). 如前所述，这种方法在 FCoE 和 RoCE 环境中的实施和采用并不常见。因此，本节不对其进行解释。本节的重点是 RoCEv2 拥塞管理 (RCM) 路由无损以太网网络中的拥塞通知。
Solution Components
The success of the approach to congestion prevention by notifying the end devices depends upon the following factors: 通过通知终端设备来防止拥塞的方法是否成功取决于以下因素：
1. Congestion detection: A switch (known as a Congestion Point) must be able to detect the symptoms of congestion. 拥塞检测： 交换机（称为拥塞点）必须能够检测到拥塞症状。
2. Sending a notification: After detecting congestion, the switch must be able to inform the end devices about it. For this purpose, Fibre Channel switches send special frames (called FPINs) and Congestion Signals. In contrast, RoCEv2 and TCP/IP networks encode this information in the data packets by marking special bits in the headers. 发送通知： 检测到拥塞后，交换机必须能够将其通知终端设备。为此，光纤通道交换机会发送特殊帧（称为 FPIN）和拥塞信号。相比之下，RoCEv2 和 TCP/IP 网络通过在报头中标记特殊位，将此信息编码到数据包中。
3. Receiving a notification: The end devices must be able to understand the notification from the switches. This depends on the hardware and/or software capability of the end devices. 接收通知： 终端设备必须能够理解交换机发出的通知。这取决于终端设备的硬件和/或软件能力。
4. Congestion prevention action: After being notified about congestion in the network, the end devices must take preventive actions, such as reducing their rate. This is the most important component. 预防拥塞行动： 在收到网络拥塞通知后，终端设备必须采取预防措施，如降低速率。这是最重要的组成部分。
The following sections explain these components using RoCEv2 Congestion Management (RCM). 以下各节将使用 RoCEv2 拥塞管理 (RCM) 对这些组件进行说明。
RoCEv2 Transport Overview
RoCEv2 packet has an IP header and hence, it can be classified using the DSCP field in the IPv4 or IPv6 header. Refer to the earlier section on Priority Flow Control for understanding how traffic can be classified and assigned to a no-drop class in routed layer 3 networks.