[no subject]

[owner-ccamp@ops.ietf.org]

> From Jonathan.Lang@sonos.com Sun Oct 30 06:23:46 2005
Received: from [70.184.40.73] (helo=sb-exchange1.rinconnetworks.com) by
psg.com with esmtp (Exim 4.52 (FreeBSD)) id 1EW6ba-0001is-DU for
ccamp@ops.ietf.org; Sun, 30 Oct 2005 06:23:46 +0000
Subject: RE: Two Drafts for Resilience of Control Plane
From: "Jonathan Lang" <Jonathan.Lang@sonos.com>
To: <ccamp@ops.ietf.org>
Sender: owner-ccamp@ops.ietf.org
Precedence: bulk

Zafar,

<snip>
> > > > > -----Original Message-----
> > > > > From: Ong, Lyndon [mailto:Lyong@Ciena.com]
> > > > > Sent: Friday, October 28, 2005 1:28 PM
> > > > > To: ibryskin@movaz.com; dpapadimitriou@psg.com;=20
> > > > > dimitri.papadimitriou@alcatel.be
> > > > > Cc: Drake@movaz.com; Drake, John E;
> > > > dimitri.papadimitriou@alcatel.be;
> > > > Igor
> > > > > Bryskin; Zafar Ali; Kim Young Hwa; ccamp@ops.ietf.org
> > > > > Subject: RE: Two Drafts for Resilience of Control Plane
> > > > >
> > > > > Hi Igor,
> > > > >
> > > > > Are you referring to controlling the LSP while one of the
> > > > controllers is
> > > > > still down?  That would mean that no restart has yet taken
> > > > > place.
> > > > > The draft seems to be focused on providing standby control
> > > > > channels,
> > > > [JD]
> > > >
> > > > What is the utility of having standby control channels, as
> > > > opposed
> > > > to just having multiple *active* control channels?
> > > >
> > >
> > > There is no advantage, but there are numerous disadvantages.
> > > Having
> > > multiple *active* control channels is what I was referring to as
> > > let IP
> > > layer provide the  resilience for Control Plane. In fact there are
> > > disadvantages of having active and standby control channel and
> > > some
> > > other entity have to manage them. Now you have to worry about
> > > signaling
> > > and control channels to be in-sync, fast switchover, etc.
> > >
> > > My motivation is to keep is simple, but yet superior.
> > [JD]
> >
> > Years ago, when we first started on LMP, we had the notion of active
> > and standby control channels and it was George Swallow that pointed
> > out that this was a really *BadIdea*tm.
> >
>
> Good news is that "we" are consistent :-) The bad news is
> that LMP newly born RFC still have this notion [no one
> listened to George :-( ].

RFC4204 does NOT include the notion of active/standby control channels.
RFC4204 does explicitly support using multiple active control channels
for an adjacency.

> Here are some reasons why "we"
> think this way:
>
> - Redundancy/ HA via IP Layer.
> - Taken a CC OOS is simple as putting an IF to the control
> network OOS, and other operational simplicities.
> - OOS IF can be repair without impacting other links or LMP CC FSM.
> - One LMP CC FSM per "N" Egress links to control network.
> - LMP CC FSM is up as long as we have connectivity to the neighbor.
> - No need to make special arrangements to make signaling and
> control channel consistent. I.e., both LMP and RSVP Hellos
> fail only when there is no IP connectivity to the neighbor.
> - Etc.

What you describe is certainly valid within a single control network.

Best,
Jonathan