Allied Telesis x900-48FE-N User Manual
Page 30 | AlliedWare™ OS How To Note: EPSR
Ports and Recovery Times
Ports and Recovery Times
In practice, recovery time in an EPSR ring is generally between 50 and 100ms. However, it
depends on the port type, because this determines how long it takes for the port to report
that it is down and send a Link-Down message.
depends on the port type, because this determines how long it takes for the port to report
that it is down and send a Link-Down message.
The following ports report that they are down immediately or within a few milliseconds,
which leads to an EPSR recovery time of 50 to 100ms:
which leads to an EPSR recovery time of 50 to 100ms:
•
10/100M copper RJ-45 ports
•
tri-speed copper RJ-45 ports operating at 10 or 100M
•
fiber 1000M ports
•
10G ports
However, for tri-speed copper RJ-45 ports operating at 1000M, there is a short delay—
either 350ms or 750ms—before the port reports that it is down. This is because the IEEE
standard for 1000BASE-T specifies that a port must wait for a certain length of time after a
link goes down before it decides that the link is actually down (see Section 40.4.5.2 of
IEEE Std 802.3-2002). The length of the wait depends on whether the 1000BASE-T port is
“master” or “slave” end of the link (“master” and “slave” are determined when the port
autonegotiates and are not related to the master node of EPSR). If a 1000BASE-T port is the
master the wait is 750ms; if it is the slave, the wait is 350ms.
either 350ms or 750ms—before the port reports that it is down. This is because the IEEE
standard for 1000BASE-T specifies that a port must wait for a certain length of time after a
link goes down before it decides that the link is actually down (see Section 40.4.5.2 of
IEEE Std 802.3-2002). The length of the wait depends on whether the 1000BASE-T port is
“master” or “slave” end of the link (“master” and “slave” are determined when the port
autonegotiates and are not related to the master node of EPSR). If a 1000BASE-T port is the
master the wait is 750ms; if it is the slave, the wait is 350ms.
This means that if a 1000M copper
link goes down between two
transit nodes, EPSR recovers after
approximately 350ms. The EPSR
nodes at both ends of the broken
link send a Link-Down message
when they detect that the link has
gone down. As the diagram shows,
the node at the slave end of the
link sends a Link-Down message in
350ms. The node at the master
end does not send a Link-Down
message until 750ms have passed,
but by then the EPSR master node
has already handled the first Link-
Down message. You can see the
messages in the debugging output
in
link goes down between two
transit nodes, EPSR recovers after
approximately 350ms. The EPSR
nodes at both ends of the broken
link send a Link-Down message
when they detect that the link has
gone down. As the diagram shows,
the node at the slave end of the
link sends a Link-Down message in
350ms. The node at the master
end does not send a Link-Down
message until 750ms have passed,
but by then the EPSR master node
has already handled the first Link-
Down message. You can see the
messages in the debugging output
in
For almost all networks, this slight delay in recovery has no practical effect. For networks
with extremely stringent failover requirements, we recommend using fiber 1000M ports
instead of copper.
with extremely stringent failover requirements, we recommend using fiber 1000M ports
instead of copper.
Master
Node
Transit
Node
Transit
Node
Transit
Node
epsr-copper
slave end
of link
master end
of link
Link-Down
after 750ms
Link-Down
after 350ms
1
2