Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

RE: [RPRWG] Phy Layer question




At 03:29 PM 4/23/2001 -0700, Karighattam, Vasan wrote:
>The debate is whether we should /
>should not ignore the LOS, LOF,
>AIS-L, etc alarms (through SF) from sonet and replace them with new RPR
>alarms.
>
>Vasan

Vasan, others,

This thread (sorry for jumping in late) has articulated a need for 
separation of RPR and physical layer (phy) and a potential need 
for using both RPR and phy triggers ("alarms") for protection. 

One comment to add would be that we actually may need BOTH triggers 
as they can be used to serve different purposes.

Chapter 9 of RFC 2892 touches on some issues regarding Packet Rings
over a sample phy (SONET), in particular coexistence of SONET and 
Packet Ring protection (IPS in this example). While it may be tempting
to state that phy (e.g. SONET, optical etc) protection will not be used 
when RPR protection is used such statement is not enforceable, 
as phy and RPR domains may be independently owned/controlled. 

As a result, a given span of RPR may also have a phy protection
provided (for example it may traverse a SONET ring). The RFC 
mentions holding off RPR trigger for e.g. 100ms to allow the SONET
protection to take place (and to trigger Packet Ring protection
only if SONET fails).

An additional improvement would be to *differentiate* between phy and
end-to-end/RPR triggers to immediately trigger RPR protection if phy
LOS/LOF/AIS-L is detected. (Note that the link between RPR node and
a SONET ADM which is on a SONET ring would not be protected
by the SONET ring). Only the triggers across the end-to-end RPR span,
like RPR keep-alive (or SONET AIS-P) would be delayed by
the above mentioned holdoff-time. (I hope it's clear without a diagram,
if not pls let me know :-)

So in summary I think that .17 needs to consider both local phy failure
alarms and the "across-the-span end-to-end" alarms (which in the 
absence of appropriate phy alarms may have to be provided by RPR
specific mechanism)

thanks

George