Chapter 8 - 25. Congestion Management in TCP Storage Networks

这篇具有很好参考价值的文章主要介绍了Chapter 8 - 25. Congestion Management in TCP Storage Networks。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。

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, maybe an overreaction leading to underutilization of the links. Also, ECN conveys only the presence of congestion. It does not convey the extent or severity of congestion. 带有 ECN 的标准 TCP 可检测到拥塞的存在,并通过减少拥塞窗口(就像丢弃数据包一样)对其做出反应。在数据中心环境中,链路容量大,端到端延迟低,标准 TCP 拥塞控制机制(如将拥塞窗口减半)可能会反应过度,导致链路利用率不足。此外,ECN 只表示存在拥塞。它并不表示拥塞的范围或严重程度。

Hence, like TCP implementation on Cisco FCIP switches, modified TCP implementations may be available on the end devices for intra-data center communication. Data Center TCP (DCTCP) is one such example. It is described in an information RFC 8257, but it is not standardized. 因此,与思科 FCIP 交换机上的 TCP 实现一样,终端设备上也可以使用修改过的 TCP 实现,以实现数据中心内部的通信。数据中心 TCP (DCTCP) 就是这样一个例子。信息 RFC 8257 对其进行了描述,但并未标准化。

DCTCP changes traditional ECN processing by estimating the fraction of bytes that encounter congestion rather than simply detecting that some congestion has occurred. DCTCP then scales the TCP congestion window based on this estimate. This method achieves high-burst tolerance, low latency, and high throughput within a data center network. DCTCP 改变了传统的 ECN 处理方式,它估算遇到拥塞的字节数,而不是简单地检测是否发生了拥塞。然后,DCTCP 根据这一估计值调整 TCP 拥塞窗口。这种方法可在数据中心网络内实现高突发容错、低延迟和高吞吐量。

The only difference between a DCTCP sender and a standard TCP sender is in how each reacts to receiving an ACK with the ECN-Echo flag set. This is explained earlier in the section on Explicit Congestion Notification in TCP/IP Networks. Other features of TCP such as slow start, additive increase in congestion avoidance, or recovery from packet loss are left unchanged. While standard TCP reduces its window size by half, DCTCP reduces the window only slightly during low severity of congestion. When congestion is severe, DCTCP reduces the window to half, just like standard TCP. DCTCP 发送方与标准 TCP 发送方的唯一区别在于,各自如何对收到设置了 ECN-Echo 标记的 ACK 作出反应。这一点在前面 "TCP/IP 网络中的显式拥塞通知 "一节中已有解释。TCP 的其他功能,如慢速启动、拥塞避免的累加增加或数据包丢失后的恢复,则保持不变。标准 TCP 会将窗口大小减半,而 DCTCP 只在拥塞严重程度较低时稍微减小窗口。当拥塞严重时,DCTCP 会像标准 TCP 一样将窗口减半。

Although modified TCP implementations, such as DCTCP, may deliver better performance, their use must be carefully analyzed because of the following reasons: 虽然修改后的 TCP 实现(如 DCTCP)可以提供更好的性能,但由于以下原因,必须对其使用进行仔细分析:

1. Configuration overhead: DCTCP or similar modified TCP implementations may not be enabled by default. Hence, using them involves a configuration overhead of enabling them on all devices. DCTCP 或类似的修改 TCP 实现可能不会在默认情况下启用。因此,使用它们需要在所有设备上启用这些配置。

2. Expertise: Deploying DCTCP or similar mo文章来源地址https://www.toymoban.com/news/detail-842782.html

到了这里,关于Chapter 8 - 25. Congestion Management in TCP Storage Networks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 - 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(35)
  • Chapter 8 - 14. Congestion Management in TCP Storage Networks

    Queue Utilization Considerations TCP/IP networks typically aim for non-empty queues so that the queues always have packets for transmission without keeping the link idle. This achieves a high transmission rate and returns on investments (cost per Gbps). TCP/IP 网络通常以非空队列为目标,使队列中始终有数据包可供传输,而不会使链路

    2024年02月21日
    浏览(37)
  • Chapter 8 - 11. Congestion Management in TCP Storage Networks

    Congestion Notification in TCP Storage Networks Congestion Notification in TCP storage networks is achieved by Explicit Congestion Notification (ECN), which is standardized by RFC 3168. As explained earlier, iSCSI and NVMe/TCP layers remain unaware of any such mechanisms. TCP 存储网络中的拥塞通知是通过显式拥塞通知 (ECN) 实现的, RFC 31

    2024年02月20日
    浏览(33)
  • Chapter 8 - 12. Congestion Management in TCP Storage Networks

    Comparison with RoCEv2 Networks RoCEv2 Congestion Management (explained in  Chapter 7 ) uses the same mechanism as ECN in TCP/IP networks. Compare Figure 8-103 with  Figure 7-17  in  Chapter 7 . You can notice that Steps 1 to 4 are the same. The difference is after a destination receives a CE-marked IP packet. RoCEv2 拥塞管理(在第 7 章中解释)

    2024年02月20日
    浏览(37)
  • Chapter 7 - 14. Congestion Management in Ethernet Storage Networks以太网存储网络的拥塞管理

    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。

    2024年01月22日
    浏览(47)
  • Chapter 7 - 15. Congestion Management in Ethernet Storage Networks以太网存储网络的拥塞管理

    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’ notif

    2024年01月22日
    浏览(51)
  • 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(44)
  • 论文笔记-Authenticated Keyword Search in Scalable Hybrid-Storage Blockchains

    混合存储模型: 只有少量meta-data(加密哈希)存在链上,原始数据外包给链下的存储服务商 提出了一个新的ADS 1.首先提出了抑制默克尔倒置(Merkle inv)索引,该索引仅在链上维护部分 ADS 结构,可以使用对数加密证明进行安全更新。 2.提出了一个变色龙倒置(Chameleon inv)索

    2024年01月18日
    浏览(42)
  • 解决RabbitMQ报错Stats in management UI are disabled on this node

    linux 部署 rabbitmq后,打开rabbitmq管理界面。点击channels,会报错 cd /etc/rabbitmq/conf.d/

    2024年02月10日
    浏览(44)
  • 【阅读笔记】Blockchain management and ML adaptation for IoT environment in 5G and beyond ...

    本文是一篇CCF C类文章,作者来自印度旁遮普邦帕蒂拉塔帕工程技术学院计算机科学与工程系 🙋‍♂️张同学 📧zhangruiyuan@zju.edu.cn 有问题请联系我~ 5G 和 B5G有什么区别? 答:5G主要解决了我们熟悉的高清视频、传输速率等问题;而B5G(Beyond-5G)将解决一些应用场景与技术的完

    2023年04月09日
    浏览(50)

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