Chapter 7 - 14. Congestion Management in Ethernet Storage Networks以太网存储网络的拥塞管理

这篇具有很好参考价值的文章主要介绍了Chapter 7 - 14. Congestion Management in Ethernet Storage Networks以太网存储网络的拥塞管理。希望对大家有所帮助。如果存在错误或未考虑完全的地方,请大家不吝赐教,您也可以点击"举报违法"按钮提交疑问。

PFC Watchdog

PFC watchdog works similarly to Pause timeout, but it only drops the traffic in the queue that is unable to transmit continuously for a timeout duration because of receiving PFC Pause frames. PFC 进程看门狗的工作原理与暂停超时类似,但它只会丢弃队列中因收到 PFC 暂停帧而无法在超时时间内连续传输的流量。

Depending upon the implementation, the PFC watchdog can trigger the following actions: 根据实施情况,PFC 看门狗可触发以下操作:

1. Port flap or shutdown: When the PFC watchdog is detected on a queue, the port is flapped or shut down, which affects all the traffic classes on this port. This is the same action as the pause timeout feature. Some people call it a disruptive watchdog. 端口翻转或关闭: 当检测到队列上存在 PFC 进程看门狗时,端口将被翻转或关闭,从而影响该端口上的所有流量类别。这与暂停超时功能的作用相同。有人称其为破坏性看门狗。

2. Alert only: When the PFC watchdog is detected, it generates an alert, such as Syslog, but does not take any further action of dropping the traffic. Some people call it a log-only watchdog. 仅警报: 当检测到 PFC 看门狗时,它会生成警报(如 Syslog),但不会采取任何进一步措施来丢弃流量。有人称其为日志看门狗。

3. Shutdown the queue: This is the most common implementation at the time of this writing. When the PFC watchdog is detected in a no-drop queue, the following actions are taken: 关闭队列: 这是本文撰写时最常见的实现方式。当检测到 PFC 看门狗处于无损队列中时,将采取以下操作:

a. All frames in that queue are dropped (discarded). 该队列中的所有帧都会被丢弃。

b. As long as this queue remains in an Rx Pause state, all newly arriving frames on other ports on this switch that are destined to go out of this queue are immediately dropped. 只要该队列仍处于 Rx Pause(接收暂停)状态,该交换机其他端口上所有新到达的、注定要从该队列流出的帧都会被立即丢弃。

c. All ingress traffic on this port that belongs to the same traffic class (as the egress no-drop class) is also dropped. All ingress Pause frames are dropped as well. This is a unique differentiator of the PFC watchdog. Recall that Pause timeout (available on FCoE ports on Cisco MDS, Nexus, and UCS) and no-credit-drop timeout (available on FC ports on Cisco MDS switches) do not affect ingress traffic on a port. The argument for not dropping the ingress traffic by these features is that congestion is directional, hence the reverse traffic should not be affected. In contrast, the argument for dropping the ingress traffic by PFC watchdog is that applications running on a slow-drain device would not benefit just by unidirectional traffic (egress traffic for slow-drain device which is the same as the ingress traffic for switchport where PFC watchdog is enabled). In fact, egress traffic from a slow drain device may increase the duration and severity of congestion, as explained earlier in the section on Correlating I/O Operations, Traffic Patterns, and Network Congestion. 该端口上属于相同流量类别(如出口无损类别)的所有入口流量也会被丢弃。所有入口暂停帧也会被丢弃。这是 PFC 看门狗的独特区别。请注意,暂停超时(在 Cisco MDSNexus UCS FCoE 端口上可用)和无损信元超时(在 Cisco MDS 交换机的 FC 端口上可用)不会影响端口上的入口流量。这些功能无损入口流量的理由是,拥塞是定向的,因此反向流量不应受到影响。与此相反,PFC 看门狗拒绝接收入口流量的理由是,在慢排空设备上运行的应用程序不会因为单向流量(慢排空设备的出口流量与启用了 PFC 看门狗的交换端口的入口流量相同)而受益。事实上,如前面 "I/O 操作、流量模式和网络拥塞的相关性 "一节所述,慢排空设备的出口流量可能会增加拥塞的持续时间和严重程度。

The exact implementation of PFC watchdog feature may differ based on the type of switch. For example: PFC 看门狗功能的具体实现可能因开关类型而异。例如

1. Cisco MDS Switches: PFC watchdog is not available on Cisco MDS switches at the time of this writing. Using the Pause timeout feature is a valid alternative because the MDS switches use only a single no-drop class for only FCoE traffic. Cisco MDS 交换机: 在撰写本文时,Cisco MDS 交换机上不提供 PFC 看门狗功能。使用暂停超时功能是一种有效的替代方法,因为 MDS 交换机只对 FCoE 流量使用单一的无损类。

2. Cisco Nexus Switches: PFC watchdog is available on Cisco Nexus 9000 and Nexus 3000 switches. The support and action (port flap, alert only, queue shutdown) depend on the model type and NX-OS version. Refer to the release notes for the latest information. Cisco Nexus 交换机: Cisco Nexus 9000 Nexus 3000 交换机上提供 PFC 看门狗功能。支持和操作(端口翻转、仅警报、队列关闭)取决于型号类型和 NX-OS 版本。有关最新信息,请参阅发布说明。

3. Cisco UCS: Cisco UCS servers support PFC watchdog in UCS Manager release 4.2 onwards. Pause timeout and PFC watchdog are mutually exclusive. In later releases, the PFC watchdog was enabled by default. 思科 UCS Cisco UCS 服务器在 4.2 版以后的 UCS 管理器中支持 PFC 看门狗。暂停超时和 PFC 进程监控相互排斥。在以后的版本中,PFC 看门狗默认已启用。

The following are the configuration details of PFC watchdog on Cisco Nexus 9000 switches as of NX-OS 10.3(1). 以下是 NX-OS 10.3(1) <文章来源地址https://www.toymoban.com/news/detail-816242.html

到了这里,关于Chapter 7 - 14. Congestion Management in Ethernet Storage Networks以太网存储网络的拥塞管理的文章就介绍完了。如果您还想了解更多内容,请在右上角搜索TOY模板网以前的文章或继续浏览下面的相关文章,希望大家以后多多支持TOY模板网!

本文来自互联网用户投稿,该文观点仅代表作者本人,不代表本站立场。本站仅提供信息存储空间服务,不拥有所有权,不承担相关法律责任。如若转载,请注明出处: 如若内容造成侵权/违法违规/事实不符,请点击违法举报进行投诉反馈,一经查实,立即删除!

领支付宝红包 赞助服务器费用

相关文章

  • Chapter 8 - 23. Congestion Management in TCP Storage Networks

    Configuring versus Operating a Shared Storage Network Another important consideration is the operational state of a network. Just because a network is configured for sharing, doesn’t necessarily mean that storage and other traffic flow through it simultaneously. For example, assume you configured a network with 50% bandwidth allocated to iSCSI and NVMe/TCP

    2024年04月09日
    浏览(38)
  • Chapter 8 - 25. Congestion Management in TCP Storage Networks

    Modified TCP Implementations Standard TCP with ECN detects the presence of congestion and reacts to it by reducing the congestion window as if packets had been dropped. In data center environments, where the links are of high capacity and end-to-end delay is low, the standard TCP congestion control mechanisms, such as reducing the congestion window by half,

    2024年03月23日
    浏览(35)
  • Chapter 8 - 24. Congestion Management in TCP Storage Networks

    iSCSI and NVMe/TCP with VXLAN Virtual Extensible LAN (VXLAN) extends Layer 2 domains over a Layer 3 data center network. The end devices do not know that their traffic is passing through VXLAN tunnels. Also, the traffic classification based on the DSCP field in the IP header and ECN bits are preserved in VXLAN. Because of these reasons, everything that has b

    2024年03月12日
    浏览(38)
  • Chapter 8 - 18. Congestion Management in TCP Storage Networks

    Comparison with Lossless Networks Recall that in Fibre Channel fabrics (explained in  Chapter 3 , “ Detecting Congestion in Fibre Channel Fabrics ”) and lossless Ethernet networks (explained in  Chapter 7 ), when the source of congestion is within an end-device, these devices are called slow-drain devices and they are detected using the metrics of the

    2024年03月14日
    浏览(48)
  • Chapter 8 - 16. Congestion Management in TCP Storage Networks

    Active Queue Management As previously mentioned, dropping or marking schemes for packets that are waiting in a queue can significantly influence TCP’s behavior on the end devices. These schemes are called Active Queue Management (AQM). 如前所述,针对在队列中等待的数据包的丢弃或标记方案会极大地影响 TCP 在终端设备上的行为

    2024年02月20日
    浏览(37)
  • Chapter 8 - 13. Congestion Management in TCP Storage Networks

    Switch Buffer Management Recall that during network congestion, a TCP sender relies on the following events for reducing its transmission rate. 回想一下,在网络拥塞期间, TCP 发送端依靠以下事件来降低传输速率。 1. When the sender detects packet loss because of the following reasons. 当发送方因以下原因检测到数据包丢失

    2024年02月21日
    浏览(39)
  • chapter14:springboot与安全

    Spring Boot与安全视频 Spring Security, shiro等安全框架。主要功能是”认证“和”授权“,或者说是访问控制。 认证 (Authentication)是建立在一个声明主体的过程(一个主体一般指用户,设备或一些可以在你的应用程序中执行动作的其他系统)。 授权 (Authorization)指确定一个主

    2024年02月14日
    浏览(32)
  • veth(Virtual Ethernet)和eth0(Ethernet 0)IN Linux

    veth(Virtual Ethernet)和eth0(Ethernet 0)是两个网络设备名称。 veth是一对虚拟以太网设备,一对veth设备通过一个虚拟链路连接在一起。它们通常用于创建类似网桥或隧道的网络拓扑结构。其中一个设备可以被看作是虚拟机或容器的网络接口,另一个设备则连接到主机的网络栈。

    2024年02月13日
    浏览(35)
  • 论文笔记——Influence Maximization in Undirected Networks

    好久没发paper笔记了,这篇比较偏理论,可能边看边记比较高效一些,仅作为个人笔记,如有解读不到的还请包涵。这篇paper的贡献有两个,首先是证明了在无向图中使用greedy可以突破 1 − 1 / e 1-1/e 1 − 1/ e 的barrier(也就是greedy在无向图上会更强),达到 1 − 1 / e + c 1-1/e+c

    2024年02月15日
    浏览(42)
  • Solidity内存布局介绍 Layout in Memory and Storage

    Solidity reserves four 32-byte slots, with specific byte ranges (inclusive of endpoints) being used as follows: 0x00 - 0x3f (64 bytes): scratch space(暂存空间) for hashing methods  Hash方法的暂存空间. 0x40 - 0x5f (32 bytes): currently allocated memory size (aka. free memory pointer) 0x60 - 0x7f (32 bytes): zero slot Solidity保留了四个32字节的

    2024年02月16日
    浏览(46)

觉得文章有用就打赏一下文章作者

支付宝扫一扫打赏

博客赞助

微信扫一扫打赏

请作者喝杯咖啡吧~博客赞助

支付宝扫一扫领取红包,优惠每天领

二维码1

领取红包

二维码2

领红包