Chapter 7 - 6. Congestion Management in Ethernet Storage Networks以太网存储网络的拥塞管理

Dedicated and Converged Ethernet Network专用和融合以太网网络

Just because a network is configured as a converged Ethernet network (lossy and lossless traffic), doesn’t necessarily mean that lossy and lossless traffic runs on it simultaneously. For example, assume you configured a network with 50% bandwidth allocated to lossless traffic and 50% bandwidth for other traffic. If there is no lossless traffic at that time, at the data layer this network is no different than any other lossy Ethernet network. On the other hand, if there is no traffic in the lossy classes, this network operates like a dedicated lossless network. In other words, it is “configured” as a converged network but “operating” as a dedicated network. 将网络配置为融合以太网网络(有损和无损流量)并不一定意味着有损和无损流量可以同时运行。例如,假设你配置了一个网络,50% 的带宽分配给无损流量,50% 的带宽分配给其他流量。如果当时没有无损流量,那么在数据层,该网络与其他有损以太网网络没有任何区别。另一方面,如果无损类中没有流量,该网络就会像专用无损网络一样运行。换句话说,它 "配置 "为融合网络,但 "运行 "为专用网络。

Consider the following examples. 请看下面的例子。

1. Cisco UCS Servers: Cisco UCS servers use converged Ethernet for carrying lossless (Fibre Channel and RoCE) and lossy (TCP/IP) traffic on the same links. But if no server uses storage via Fibre Channel or RoCE, internal links will never report Pause frames, although the configuration for converged Ethernet remains applied on them. Refer to Chapter 9, “Congestion Management in Cisco UCS Servers,” for more details on Cisco UCS servers. Cisco UCS 服务器使用融合以太网在同一链路上传输无损(光纤通道和 RoCE)和有损(TCP/IP)流量。但是,如果没有服务器通过光纤通道或 RoCE 使用存储,则内部链路将永远不会报告 "暂停 "帧,尽管聚合以太网的配置仍应用于这些链路。有关 Cisco UCS 服务器的更多详细信息,请参阅第 9 "Cisco UCS 服务器中的拥塞管理"

2. FCoE on Cisco MDS Switches: Although the FCoE ports on Cisco MDS switches are configured for converged Ethernet, they only handle lossless traffic, and do not send/receive traffic in lossy classes. 虽然 Cisco MDS 交换机上的 FCoE 端口是为融合以太网配置的,但它们只能处理无损流量,不能发送/接收有损类流量。

A key point to understand is that even the dedicated lossless networks are configured the same as the converged networks. The configuration of traffic classification, bandwidth guarantee, flow control, etc. as explained in the previous section is applied in both types of networks. However, traffic pattern makes one different from the other. For congestion detection and troubleshooting, verifying configuration is the first step, which remains the same for both types of networks. The next step is to focus on the traffic pattern which depends on if the network carries only lossless traffic (dedicated) or if the network carries lossless and lossy traffic simultaneously (converged). 需要了解的关键一点是,即使是专用无损网络,其配置也与融合网络相同。上一节中解释的流量分类、带宽保证、流量控制等配置在这两类网络中都适用。不过,流量模式使两者有所不同。对于拥塞检测和故障排除,第一步是验证配置,这对两种类型的网络都是一样的。下一步是关注流量模式,这取决于网络是只传输无损流量(专用),还是同时传输无损和有损流量(融合)。

Understanding Congestion in Lossless Ethernet Networks了解无损以太网网络中的拥塞问题

Lossless Ethernet networks are prone to similar types of congestion as Fibre Channel fabrics because both use flow-control between directly connected devices. Also, this congestion spreads toward the source of the traffic victimizing many other devices that share the same network paths and traffic class. 无损以太网网络容易出现与光纤通道结构类似的拥塞,因为两者都在直接连接的设备之间使用流量控制。而且,这种拥塞会向流量源蔓延,使许多共享相同网络路径和流量类别的其他设备受害。

你可能感兴趣的:(Linux,kernel,linux,kernel,网络)