|
The
primary issue to keep in mind when connecting other
equipment to a Shore Microsystems Ethernet Link
Protector is the issue of link speed and duplex
configuration. Although some SM-series units
(e.g. the SM-2501) will allow auto-negotiation of
link speed and duplexing, it is usually a better
solution to "hard code" the connected
equipment to the speed and duplex settings desired
for the link. Presetting these parameters
allows quicker link setup when a switchover event
takes place. For example, for the SM-2501
(a single port 100BaseT unit), it is best to hardcode
all connected ports (protected, primary, and backup)
to 100Mbps and set the duplex configuration to either
full or half on all attached ports. This easily
accomplished on many systems with simple port configuration
commands or NIC driver initialization commands.
Also
see the Spanning Tree/MAC FAQ for additional information.
|
Shore Microsystems
Link Protector®s monitor a primary and backup connection
to the external network. This usually translates
to connections to primary and backup switches which
operate on the MAC address of the protected (server)
port. In normal operation, the protected port
is connected to the primary port and thus the primary
switch will learn and advertise the presence of
this MAC address to the rest of the network.
One advantage of the SM-2XXX series is that it is
only possible for the protected port to be connected
to either the primary or backup switches at any
one time (never both) and thus prevent routing loops.
However, in the event that a primary port fails,
it is necessary for the backup switch to relearn
the MAC address of the server port to which it is
now attached.
To
guarantee quick relearning of the MAC address on
the backup switch port, it is advisable to run some
type of process on the protected port device (typically
a server) that will regularly launch Ethernet frames
toward the SM-2XXX unit. For example, a continuous "ping"
process is simple and effective for this function.
The pings should be launched toward a IP address
located somewhere on or beyond the primary or backup
switches.
To
further speed the MAC relearning process, several
switch vendors provide improved methods for propagating
MAC address information through the network.
For example, Cisco Systems products implement the "Portfast"
method where the normal spanning tree requirements
are curtailed and a port can immediately transition
from a "down" state to the "forwarding"
state. With Portfast, it is possible to switch
between primary and backup ports without losing
frames. Other switch vendors provide similar
capabilities.
|
Soft failures are defined
as those sorts of failures in which the low-level
link is operational but some higher level function
is not operable. For example, if a port card
in a switch has a firmware crash, the link may still
appear to be functioning normally but no useful
data transfer can occur. This is possible
since the hardware physical (PHY) chip or chips
may still be powered up and "happy" even
though the firmware crash prevents the overall port
card from operating correctly.
In
this case, the SM-2XXX units will not respond to
this type of failure since no link failure has occurred.
This sort of problem, although less likely than
a cable/wiring problem (which will cause the SM-2XXX
to automatically respond to the failure), can be
handled by external management functionality.
One way to implement this sort of functionality
is for a network manager application to routinely
poll the protected device and in the event that
the polling is unsuccessful, use the SNMP management
features to force a switchover event to the backup
port. This capability can also be implemented
with a simple shell script.
SM-27XX Series Link Protector®s
offer the additional capability of idle packet switching.
With this feature, a programmable threshold of received
frames can trigger a switchover event. For
example, if a frame is no received in 2 seconds,
the link can be considered as "failed"
even though the link is still "up" in
the Layer 1 sense.
|
