[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference netcad::hub_mgnt

Title:DEChub/HUBwatch/PROBEwatch CONFERENCE
Notice:Firmware -2, Doc -3, Power -4, HW kits -5, firm load -6&7
Moderator:NETCAD::COLELLADT
Created:Wed Nov 13 1991
Last Modified:Fri Jun 06 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:4455
Total number of notes:16761

676.0. "Redundency Explanation Needed" by HGRNCC::FARADAYCHONG (Faraday Chong@hgo 852-8053590) Tue Feb 01 1994 11:34

    In the US Network Academy, repeaters with software redundent capability
    was introduced, can someone from engineering explain what config will
    be supported and what is not so that we can propose high availability
    network today that can take advantage of the redundent capability as it
    becomes available?
    
    Faraday
T.RTitleUserPersonal
Name
DateLines
676.1Redundancy discussion (from DECrepeater doc's)KALI::FORSBERGNIPG, Hub Products GroupFri Jun 24 1994 15:58239
    Sorry about the delay.  The following text is adapted from the
    DECrepeater 90FS (and 900FP) documentation.  The first few sections
    define terminology; the supported configurations follow.
    
Redundant Fiber-Optic Link Overview
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Redundant ports are usually configured between two or more repeaters to provide 
a standby link in case the active link fails.  If the active link fails, the 
standby link is automatically activated to handle traffic between the repeaters.

The basic redundant configuration consists of a  pair of master ports and two 
remote ports.  The master ports control which of the two lines is active and 
which is standby, and, therefore, the master ports are referred to as a master 
pair. 

The master ports can be connected to two remote ports on the same repeater 
or on different repeaters, but the remote ports must be on the same LAN (or 
extended LAN).

NOTE:     
This redundant configuration does not result in a repeater loop, because only 
one of the links is active.

The remote ports can be one of two types: a responder port or a non-responder 
port.  A responder port provides full fault detection because it signals the 
master if it detects a failure.  A non-responder port provides only partial 
fault detection because it does not signal the master if a failure occurs.

NOTE:     
The DECrepeater 90FS and the DECrepeater 900FP are the only current 
DECrepeaters that can be configured as redundant port repeaters.

Redundant Fiber-Optic Link Components
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

The following components make up redundant fiber-optic link configurations:

o Redundant-Link Pair

The RLP is a pair of fiber-optic links that provides backup communication in 
the event of transmitter, receiver, or cable failures.

o Master

The master is the DECrepeater that contains the master control algorithm (MCA) 
for the ports in the RLP. This algorithm controls the Link Enable signal that 
goes to each of two ports in the RLP. The master uses this signal to enable 
one port at a time.

o Responder Port

The RLP can contain two responder ports (referred to as responders) on the 
remote end of the links from the master ports. The responders are
independent  of each other, but reside on the same LAN (or extended LAN) as
each other.  A responder can signal the master when it detects a failure on
its receive  portion of the link. The responder does this by disabling the
Link Idle signal  on its transmit port until the receive portion of the
link is operational again.

o Non-Responder Port

Any standard 10BaseFL device ports can be non-responder ports (referred to 
as non-responders). Unlike a responder, a non-responder cannot signal the
master  when there are receive link failures on its end of the link. If you
define  an RLP with non-responders, only partial fault detection is
provided on the  RLP.

o Primary Link

Defined at initialization time by the MCA, the primary link is initially
the  active link providing Ethernet communications. Activity defaults to
the primary  link after a reset condition.

o Secondary Link

The secondary link is the other link of the RLP to which the master fails
over  if there is a failure in the primary link.

o Active Link

The active link is the link that is currently enabled by the master to carry 
Ethernet traffic.

o Standby Link

The standby link is the link that is currently disabled by the master and
cannot  carry Ethernet traffic. This port is held in readiness as a backup
link in  case the active link fails. On a functional link, the responder
can still place  traffic on the link, but there is no traffic from the
master on that link.

o Master Control Algorithm

The Master Control Algorithm (MCA) enables only one link of the RLP at any 
time and controls the RLP as follows: 

1.  The MCA enables the primary link at initialization time (assuming that
    there  are no failures on that link). This causes the primary link to
    become active.

2.  The MCA monitors the primary link status for failures.

3.  When the MCA detects a failure of the active link, the MCA:

    a.  Disables the active link.

    b.  Enables the standby link.

The active link always remains enabled (active) until a failure is detected 
on that link.

Redundant Fiber-Optic Link Operation
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
When you initially configure the DECrepeater for redundant operation, 
the primary port is the active port and the secondary port is the standby
port.  In operation, the active port communicates with its remote responder
port until  the master detects a link failure or until the port is
internally partitioned.  If the standby port is operational, the primary
port is disabled, and the secondary  port is enabled as the active port.
The failover typically occurs within 4  to 8 ms (15 ms maximum).

Once the primary link is repaired, it becomes the standby link. If the
repeater  detects a failure in the (now active) secondary link, or if the
port is internally  partitioned, failover occurs back to the primary port.


A link failure can be caused by any of the following occurrences:

o Loss of carrier (light in a fiber-optic device)

o Auto-partitioning of the port

Simple Redundant Link Configuration
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

The DECrepeater firmware allows the configuration of a redundant
fiber-optic  Ethernet link from multiple ports. In its simplest form, a
redundant link consists  of a pair of master ports that reside on one
module, each connected to independent  ports on a remote module. These
remote ports may be responder  ports or non-responder ports.

Once these ports are properly configured using HUBwatch and connected
together,  one connection is the active link and the other is the standby
link.

Complex Redundant Link Configuration
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

The DECrepeater also supports a more complex redundant link topology. 
In general, the redundant link consists of a pair of master ports on one
module,  each connected to enabled remote ports. The remote ports must
reside on the  same LAN (or extended LAN), but can be on separate modules.
The remote ports can be responders, non-responders,  or a combination of
the two. If your configuration includes a combination of  a responder and a
non-responder, configure the responder as the primary link,  and the
non-responder as the secondary link.

Partial Fault Detection
~~~~~~~~~~~~~~~~~~~~~~~
Partial fault detection causes the active master port to fail over to the
standby  port if the active port detects any of the following
circumstances:

o A link failure in in the receive portion of the active port

o An internally detected failure or partitioning of the active port

    - An excessive number of collisions

    - An excessively long collision

    - Received jabbering

    - Internally detected MAU failure

When the DECrepeater fiber-optic ports are connected  to a
non-responder  repeater, such as the DECrepeater 90FL, 90FA, or any IEEE
802.3 10BaseFL repeater port, partial fault detection is available.

Partial fault detection adequately detects most failure conditions, but
cannot  detect a failure in which the remote port cannot receive
transmissions. That  level of failure detection requires full fault
detection.

Full Fault Detection
~~~~~~~~~~~~~~~~~~~~

The DECrepeater supports full fault detection. This feature enables the 
master module to react to failures at either end of the receive or transmit 
media.

If a failure occurs in the responder's receive port, the responder signals 
its partner master port. This allows the master to switch over to the standby 
port.

When the DECrepeater fiber-optic ports are connected to another
redundant  fiber-optic repeater, such as the DECrepeater 900FP or another
DECrepeater full fault detection is available.

Table 3 shows the difference in failure detection between partial fault
detection  and full fault detection.


Table 3:   Partial Fault Detection Versus Full Fault Detection

---------------------------------------------------------------------------
Link Failure from the   |			   |
Master Port's Perception|			   |
			|			   |
    TX	       RX	| Partial Fault Detection  |  Full Fault Detection
---------------------------------------------------------------------------
Failure			|   not detected	   |     detected
			|			   |
	    Failure     |   detected		   |     detected
			|			   |
Failure	    Failure	|   detected		   |     detected
---------------------------------------------------------------------------


Recommendations for Redundant-Link Configuration
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Consider the following recommendations when configuring the DECrepeater 
for redundancy to increase fault detection and improve network reliability:

o Use responder ports (rather than non-responder ports) whenever possible.

o If you use a combination of responder ports and non-responder ports,
  assign  the responder as the primary link, and the non-responder as the
  secondary link.

o Assign the remote ports to separate modules.  This ensures that one of
  the  two remote ports is functional in the event of a remote  module
  failure.

o Whenever possible, configure primary and secondary link cables into
  separate  pathways, to avoid a single point of failure.