The MPLS WG Archive

[Ben Mack-Crane <Ben.Mack-Crane@tellabs.com>]

Dave,

I like your view.  Some comments/questions below.

Regards,
Ben

dallan@nortelnetworks.com wrote:
> 
> Ben:
> 
> Thanks for your comments. Rooting back through the other drafts, this
> discussion should really be focused around the framework draft
> <makam-mpls-recovery-frmwork-00>. To my mind the issue is whether MPLS is a
> black box and the very worst manifestation of a failure is a 'x' msec.
> interruption of packet flow out of a particular interface, OR whether we
> allow "less specific" box behaviors such as the packets keep flowing out of
> the box, but it may be a different interface. Seems to me that there are
> cases for both.

In the intiial framework we had focused on "MPLS Path Recovery," and since 
a path has fixed endpoints...  Well, there's where the blind spot came from.  
The current position of the framework, "MPLS Based Recovery," admits a 
broader interpretation and so it makes sense to incorporate an option to 
have diverse egress points for working and recovery paths.  

Any comments from other framework authors?

> 
>      <snip>
> 
> > You are correct in noting that the PML is limiting, as
> > a single egress point for the MPLS protection domain;
> > however, removing this limitation requires some knowledge
> > beyond that necessarily held by the MPLS layer.
> >
> Yes, there would have to be a path diversity test combined with some
> link-state knowledge of the over all network to ensure that the alternate
> egress point was appropriate.
> 
> > There are two extensions that I can think of based on your
> > comments (please let me know if I have misunderstood your
> > comments).
> >
> > 1) The working and recovery paths never merge as LSPs.
> > Instead, the recovery path terminates at an LER that is an
> > acceptable alternative to the LER terminating the working
> > path for the purpose of forwarding the FEC represented by
> > the working path.
> >
> > In this case, the protection domain is still within a
> > contiguous MPLS cloud
> >
>      Effectively yes. We have not discussed fast failure detection,
> propagation, and pre-planned alternate route mechanisms outside the MPLS
> space.
> 
> > (which BTW implies nothing in particular
> > about administrative domains).
> >
>      Agreed, however such things MAY exist, such as intercarrier
> boundaries inside a contiguous MPLS space. That would be my only
> observation. The boundary MAY not be technical and in the real world that is
> what frequently happens. If there is some form of pre-emption, then it is
> also useful to constrain the space to minimize the amount of traffic
> distrupted.
> 
> > The protection mechanism
> > still works with information available at the MPLS layer
> > (fault detection, protection actions, etc.).  However, selection
> > of the recovery path (outside the scope of this ID) requires
> > L3 knowledge to determine that the LER terminating the
> > recovery path is acceptable.
> >
>      I would broaden the definition as selection suggests to me explicit
> routing, whereas increasing the reliabilty of implicit path construction
> would imply intelligent pruning of options. To a certain extent,
> conservative label retention in an implicit case implies some knowledge of
> which LSPs would be more optimal than others, and sufficient information
> should be available at the overall L3 layer (not just MPLS topolocy) to make
> that determination.

Selection, in this case, was meant generically (not intended to imply explicit
routing); however, the path protection mechanism draft does not address
reroute recovery (this is covered in other IDs) and so the broadening
you suggest may be out of scope for that ID.  Is there a need to address this 
in the framework (I don' think the framework says anything in particular about 
recovery path selection options)?

> 
> > In practice this may be an attractive option, although
> > some thought may be required to apply this in an end-to-end
> > recovery design for other than best effort traffic.
> >
>      Yes, bang on!, to my mind there is two goals to this effort in
> general. First is providing explicit recovery mechanisms for specific flows
> that require specific service guaratees, and second is raising the
> reliability bar overall. You're correct in identifying best effort, as e2e
> RSVP reservations carried through such an "LSP tunnel" require explicit
> black box behaviour to avoid the interruption of re-routing on a protection
> switch (or establishing a new "deaggregator"). That is NOT the scenario that
> I think a non-PML solution adds value to.

I agree.  The diverse egress case may be paricularly useful for improving the 
reliability of best effort traffic.

> 
> > 2) The protection domain spans both MPLS hops and L3 hops.
> >
>      Not sure how this really differs from '1' except the suggestion that
> there are fast non-MPLS restoration mechanisms which require some overall
> coordination.
> 
> > Here the protection domain is not within a contiguous MPLS
> > cloud, and recovery within such a protection domain may be
> > complex.  Since the protection domain includes both MPLS and
> > L3 hops, fault detection and recovery actions may be significantly
> > different depending on where a fault occurs (and the protection
> > mechanism would require L3 knowledge in some cases).  It may
> > be difficult to guarantee uniform recovery performance across
> > the entire protection domain.
> >
> >
>      I think in general it is easiest to think of this as concatenated
> protection domains with redundancy at the domain boundary. Ultimately it is
> a form of "Per-Hop Protection" ;-)  as the lack of uniformity of mechanisms
> makes explicit guarantees difficult but the performance is the best the
> overall network can do. If we stick to that definition then there is really
> no difference between 1 and 2.

I was concerned that the broader defintion of protection domain being proposed 
implied that ONE recovery action (e.g., PSL switch) would be able
to recover from any fault in the protection domain.  I had trouble seeing
how a fault on an L3 link could be correlated with the working LSP, and the 
PSL informed.  The concept of concatenated protection domains is much
simpler, and I think this allows a narrower definition of protection
domain (uniform mechanism).  Redundancy/diversity/coordination at protection 
domain boundaries is then the interesting problem.

> 
> > The advantage of this broader definition of protection domain
> > is that it is not necessary to merge the traffic at arbitrary
> > boundaries (e.g., at LSP terminations, AS boundaries, etc.).
> > The complexity, however, makes this definition more difficult
> > to characterize, and makes defining a coherent protection
> > mechanism harder.
> >
> >
> > Both of these extensions require some L3 knowledge, but the
> > second is the only one that requires L3 knowledge in the
> > protection mechanism itself.  The original mechanism was
> > intended to work entirely within the MPLS layer.  These
> > potential extensions would broaden the scope beyond MPLS,
> > at the cost of additional complexity.
> >
>      If I get your inference, you assume that a non-MPLS protection
> mechanism requires some type of resource reservation.

Well... I wasn't trying to say anything about non-MPLS mechanisms.
We are trying to further clarify the framework terminology (a summary
has been posted to the mail list by Vas Makam) and "protection
switching" implies resource reservation.  This is not to say that
there are not recovery mechanisms that work otherwise (e.g., reroute
recovery).  The path protection mechanism ID is aimed at protection
switching, so resource reservation is assumed there, and this leads
to some desire to have a mechanism that operates simply at the MPLS 
layer.

> 
> > Perhaps it is better to live with some limitations so that
> > the mechanism is simpler (there is some value in a "tidy,
> > administrable and measurable definition of how far protection
> > extended").  Thoughts?
> >
>      My summary would be:
> 
>      1) We need to clarify overall the goals of MPLS restoration. Is it
> explicit black box behavior or are we seeking a broader base of solutions
> that dovetail into the overall network. If we are discussing a broader base,
> then a 1:1 or n:1 non-"lsp merging" solution has applicability. If we are
> "black box" only as it works for all style of protections (1+1, n:1, 1:1),
> all styles of traffic (best effort, diffserv and RSVP pinnned route) and is
> measuable, then we'll have to tolerate less optimal e2e routing and topology
> to get that. Personally, I'd like a bigger toolbox.

It makes sense to have a bigger toolbox, as long as we have a reasonable 
understanding of the tools and their appropriate use.  You have pointed out
a useful alternative to be added, and we should try to develop a clear
description for the affected IDs.

> 
>      2) I don't think we need to overload the recovery space by expanding
> the protected domain beyond MPLS, only acknowledge that we may concatenate
> protection mechanisms, they function autonomously, and may have multiple
> points of attachment. Coordination of protection across multiple transport
> technologies is out of scope if we are to make progress.

This seems reasonable.

> 
>      cheers
>      Dave
> 
>   --------------------------------------------------------------------------------
>                   Name: att1.htm
>    att1.htm       Type: Hypertext Markup Language (text/html)
>               Encoding: 8bit
>            Description: Internet HTML