openBackhaul / synchronization

Synchronization model according to ITU-T G.7721-2018
Apache License 2.0
1 stars 0 forks source link

1000BaseT Ethernet interfaces #62

Open niloda opened 8 months ago

niloda commented 8 months ago

1000BaseT electrical Ethernet interfaces implicitly create a clock loop. In accordance with the IEEE 802.3-2018 standard (§1.4.456 slave Physical Layer (PHY)), when they are configured as slaves (see rx-sync-preference of wire-interface-configuration), the transmission clock is equal to the reception clock. This feature could create problems in a ring controlled by the SSM protocol.

As shown in the following figure, the clock runs counterclockwise giving LAN-1 a higher priority than LAN-2.

image

In case of failure between Ne-2 and Ne-3, Ne-3 is synchronized by changing the clock direction between Ne-1, Ne-2 and Ne-3. The role of the LAN-2 port of Ne-2 and Ne-3 must be forced to slave so that the clock direction can be reversed.

image

Of course, for the clock direction to be consistent in the ring, the master/slave roles of the 1000BaseT Ethernet interfaces should be driven by the SSM protocol. For this purpose, some devices provide a parameter to choose whether the role of the Ethernet interface is static and defined by the interface configuration (rx-sync-preference of wire-interface-configuration) or whether it must be driven by the SSM protocol.

The current model doesn't seem to provide a parameter like that.

niloda commented 7 months ago

According to this morning's discussion, the clockDirectionDrivenBySssm parameter (boolean with default set to true) is fine for us if only added to the fcPort augmentation.

openBackhaul commented 7 months ago

Decision during the LF ONMI x-haul call on 3rd of April 2024: Two attributes to be added to the FcPort::SyncFcPortSpec class: