Technical data

TraverseEdge 2020 Applications and Engineering Guide, Chapter 3: System Applications
Release 5.0.x Turin Networks Page 3-53
The GbE PLM will transmit at line rate any properly formatted Ethernet MAC frame. These include:
• Unicast, multicast, and broadcast frames
• Bridge Protocol Data Units (BPDUs)
• Jumbo frames up to 10,240 octets long (the default maximum frame size is 1518 octets)
• Frames with VLAN tags and nested VLAN tags (enhanced GbE PLM only)
• Frames with MPLS shim headers and nested MPLS shim headers (enhanced GbE PLM only)
The response of the GbE PLM to received 802.3x PAUSE frames may be provisioned. Each interface may
be independently configured to operate in one of two modes:
1. Pass-through: The GbE PLM forwards received PAUSE frames without processing them. This
allows the two end systems to run Ethernet flow control transparently. However, the effectiveness
of the flow control will depend on the end-to-end delay, which, in turn, depends on the network
design. Service providers who do not wish to reveal important aspects of their network design to
their subscribers should advise them to disable flow control over the GbE interfaces.
2. Local response: The GbE PLM responds to a received PAUSE frame in compliance with the
802.3x procedures for processing PAUSE frames (i.e., it stops sending frames for the amount of
time specified in the PAUSE frame). The GbE PLM also generates a PAUSE frame that it sends
downstream to the egress GbE PLM. While not transparent to the two end systems, this mode of
operation ensures that the congested end system - the one that generated the PAUSE frame - will
see a more rapid response to the PAUSE frames it sends.
For a given GbE "private line," the user should configure the ingress and egress GbE interfaces to the same
mode of PAUSE frame response.
The GbE PLM supports auto negotiation, and will interoperate with end systems that auto negotiate speed,
duplex, and PAUSE frame operation. The GbE PLM will auto negotiate to 1000 Mbps, full duplex opera-
tion. During auto negotiation for PAUSE frame operation, the GbE PLM will indicate "both symmetric
PAUSE and asymmetric PAUSE toward the local device" - essentially a "don't care" that allows the end
system to determine the PAUSE frame operation on each GbE link. The user must therefore ensure that the
end systems are configured to auto negotiate to the desired PAUSE frame operation. Some guidelines:
• If the ingress and egress GbE PLMs are configured to operate in pass-through mode, and if the
user desires MAC layer transparency (i.e., all PAUSE frames pass from one end system to the
other), then both end systems should be configured to auto negotiate for symmetric operation.
While the GbE PLM will pass all PAUSE frames transparently, an end system configured for
asymmetric flow control will discard PAUSE frames received from the GbE PLM.
• If the ingress and egress GbE PLMs are configured to operate in local response mode, any
PAUSE frame configuration of the end systems will function correctly.
• If the user wants to prevent one of the GbE end systems from being flow controlled (e.g., a ser-
vice provider does not want to give the CPE the capability to throttle a service provider GbE inter-
face using PAUSE frames), then the GbE interface on that system should be configured to auto
negotiate for asymmetric PAUSE frame operation. This ensures that the GbE interface on that end
system will discard received PAUSE frames.
In addition to MAC layer transparency, the GbE PLM provides some degree of physical layer transpar-
ency. While the GbE PLM terminates the 8B/10B line coding at each GbE interface, it propagates physical
layer fault indications. Any physical layer failure - whether at the near end GbE link or within the SONET