Chapter 7 - 8. Congestion Management in Ethernet Storage Networks以太网存储网络的拥塞管理

Stomped CRC Counters

Stomped CRC counters help in finding the location of bit errors in a network that uses cut-through switches. More precisely, these counters help in finding where bit errors do not exist. Stomped CRC 计数器有助于在使用直通式交换机的网络中查找位错误的位置。更确切地说，这些计数器有助于找到不存在位错误的地方。

When a cut-through switch support stomped CRC feature, it encodes a special value in the CRC/FCS field of the corrupt frame. This is called frame stomping. As mentioned, the switch can’t drop the corrupt frame because it already starts transmitting it. But it can stomp the frame because it knows the frame is corrupted and it is yet to transmit the CRC field at the end of the frame. When the next switch detects a stomped frame, it increments only the stomped CRC counter and does not increment the CRC counter. When all these ports are compared, ports with stomped CRC error can be excluded from the investigation and the port/link with CRC counters can be investigated. The corrupt frame is eventually dropped at its destination or on a store-and-forward switch. 当直通交换机支持踩踏 CRC 功能时，它会在损坏帧的 CRC/FCS 字段中编码一个特殊值。这就是所谓的 "帧踩踏"。如前所述，交换机不能丢弃损坏的帧，因为它已经开始传输。但交换机可以踩帧，因为它知道帧已损坏，而且尚未传输帧末的 CRC 字段。当下一个交换机检测到被踩踏的帧时，它只会递增被踩踏的 CRC 计数，而不会递增 CRC 计数。对所有这些端口进行比较后，就可以排除有踩踏 CRC 错误的端口，对有 CRC 计数器的端口/链路进行调查。损坏的帧最终会在目的地或存储转发交换机上丢弃。

Example 7-9 shows Stomped CRC errors on Cisco Nexus 9000 switches and Cisco UCS servers using the NX-OS command show interface. 例 7-9 使用 NX-OS 命令 show 界面显示了 Cisco Nexus 9000 交换机和 Cisco UCS 服务器上的 Stomped CRC 错误。

Forward Error Correction

When FEC is enabled, a sender adds a few additional parity bits into the bitstream. The receiver can use these parity bits to detect and recover a limited number of bit errors. 启用 FEC 时，发送方会在比特流中增加一些额外的奇偶校验位。接收器可利用这些奇偶校验位检测和恢复有限的比特错误。

When FEC is able to recover the corrupted bits: 当 FEC 能够恢复损坏的比特时：

 FEC corrected counter increments, and经校正的 FEC 计数器增量，以及

 CRC counter does not increment because FEC already recovered the bit errors at a lower layer and bits are handed off to the framing layer as sent by the sender. CRC 计数器不会递增，因为 FEC 已在较低层恢复了比特错误，比特将按照发送方发送的方式移交给成帧层。

When FEC is unable to recover the corrupted bits: 当 FEC 无法恢复损坏的比特时：

 FEC uncorrected blocks counter increments, and    FEC 未校正块计数器增量，以及

 CRC counter may increment if the bit error is within a frame. 如果位错误发生在一个帧内，CRC 计数器可能会递增。

On Cisco Nexus 9000 switches, use the command show hardware internal tah mac hwlib show mac_errors fp-port to display FEC Correctable and FEC UnCorrectable counters. As Example 7-11 shows, this is a module level command so before using it, you must use NX-OS command attach module. 在 Cisco Nexus 9000 交换机上，使用命令 show hardware internal tah mac hwlib show mac_errors fp-port 显示 FEC Correctable 和 FEC UnCorrectable 计数器。如例 7-11 所示，这是一条模块级命令，因此使用前必须使用 NX-OS 命令 attach module。

Example 7-11 FEC Counters on Cisco Nexus 9000 switches

switch# attach module 1

module-1# show hardware internal tah mac hwlib show mac_errors fp-port 15

<snip>

MAC: HSMCPCS Err per channel:     0        1       2       3

------------------------------------------------------------

BER Count                        .....   .....   .....   .....

Err Blocks Count                 .....   .....   .....   .....

Sync Loss Count                  .....   .....   .....   .....

Block Loss Count                 .....   .....   .....   .....