The MPLS WG Archive

Cell Relay Retreat>MPLS WG Archive>month:2001-Apr> msg00049



[Date Prev][Date Next][Thread Prev][Thread Next]  
  [Date Index][Thread Index][Author Index][Subject Index]

Last Call on RSVP Label Allocation for Backup Tunnels

  • From: Bora Akyol <akyol@pluris.com>
  • Date: Tue, 3 Apr 2001 20:27:10 -0700
  • Cc: Bora Akyol <akyol@pluris.com>, Dave Cooper <dcooper@gblx.net>, mpls@UU.NET

At 7:26 PM -0700 4/3/01, Ping Pan wrote:
>Bora Akyol wrote:
>>
>>  Ping
>>
>>  At least in one implementation of APS and FR that I am familiar with, the
>>  timing to generate a switchover between the two is very close in terms of
>>  failover and in this particular implementation, APS switchover timing meets
>>  less than 50 ms.
>>
>>  So I would not immediately rule out FR timing in place of APS.
>>
>>  Bora
>
>Bora,
>
>I was pointing out that the fail-over time can be *unpredictable* at
>protocol level, where a link failure may trigger heavy processing among
>many protocols (such as BGP and RSVP). I have no doubt that if you put a
>link detection routine within link card's driver or something, the
>failover time can be quite fast.
>
>- Ping

Agreed but what I was trying to say thatthe failover time despite 
being dependent on the number of LSPs that have fast reroute enabled 
is not terribly different than SONET APS switchover times. But yes if 
you have N LSPs where N is really large the failover may take longer 
than if you had only one. In practice you will probably not have more 
than 1000 LSPs so this processing delay is not huge.

Bora