Re: [PWE3] VCCV usage

[Sriganesh Kini <[e-mail]>]

IMO, a PE cannot connect to two different PEs for the "same" PW. This
also holds true for any segment of that PW. Also, a PE cannot connect
to the same PE with different PW labels for the "same" PW. This too
holds true for any segment of that PW. It may be useful for the MS-PW
draft to clarify that. Note that this does not preclude a PW from
being a backup of another PW (possibly with different number of PW
hops).

The notion of different PW hops for any single PW due to ECMP should
not arise. Note that there may still be different number of hops in
the underlying MPLS PSN.

-- 
- Sri

On Mon, Nov 29, 2010 at 11:30 PM, Alexander Vainshtein
<[e-mail]> wrote:
> Greg, Carlos and all,
> I am probably missing something important, but I see the situation when VCCV
> Type 3 may behave differently due to different number of hops on different
> ECMP paths somewhat problematic to me.
>
> This is because the hop count for MS-PWs is the S-PE count, not the count of
> hops between one S-PE to another one.
>
> In order to traverse a different number of S-PEs due to ECMP, you would need
> a combination of at least two elements:
>
> Multiple S-PE paths being explicitly between the two T-PEs (for a given
> MS-PW). I am not even sure this is allowed by the MS-PW architecture today
> (even with fat PWs)
> S-PEs doing ECMP on PW labels.
>
> My 2c,
>
>      Sasha
>
>
> ________________________________
> From: [e-mail] [[e-mail]] On Behalf Of Greg Mirsky
> [[e-mail]]
> Sent: Tuesday, November 30, 2010 1:05 AM
> To: Carlos Pignataro
> Cc: [e-mail]; Luca Martini (lmartini); Yaakov Stein;
> [e-mail]
> Subject: Re: [PWE3] VCCV usage
>
> Dear Carlos,
> you've written:
> "If the VCCV packet takes a different path than the PW data, can this
> different path have a different number of hops and have the VCCV expire
> at a PE different than the target PE?"
> I wouldn't say that path taken by VCCV packet is different from the path of
> PW payload. If ECMP exists, VCCV packet will travel over one of available to
> PW payload paths.
> I think it is possible, though might be not realistic scenario, that number
> of segments between pair of PEs is different over different paths of ECMP.
> Regards,
> Greg
> On Mon, Nov 29, 2010 at 2:35 PM, Carlos Pignataro <[e-mail]>
> wrote:
>>
>> Hi Sri,
>>
>> Thanks for the summary. Please see inline.
>>
>> On 11/29/2010 4:10 PM, Sriganesh Kini wrote:
>> > Carlos, that is useful clarification. If I were to summarize,
>> >
>> > 1. When a PW has a CW, CC Type 1 is obvious/implicit choice.
>>
>> Or put a different way, CC Type 1 can only be used when the PW has a CW.
>> There's no implicit choice standardized though.
>>
>> > However,
>> > PW label TTL expiry should trigger examining the ACH for an VCCV
>> > packet (both S-PE and T-PE).
>>
>> A PW label TTL expiry should trigger the exception mechanism to dealing
>> with a TTL expiration; this may include examining the PW-ACH for a VCCV
>> packet (or not), and may include detecting the VCCV packet but not
>> processing it for security if not advertised.
>>
>> > 2. When PW does not have CW, CC Type 3 should be used.
>>
>> According to RFC 5085, that should be Type 2 as preferred over Type 3
>> (see S5.1.2. and S7)
>>
>> > When PWs were
>> > single segment, the 'PW demultiplexer' TTL was defined as
>> > application-specific (RFC 3985). With MS-PW it is important for it to
>> > have consistent TTL decrement operations and treat TTL expiry (i.e.
>> > TTL=1) packets as VCCV packets (both at S-PE and T-PE). This requires
>> > the number of PW hops to be known to execute OAM targeted to a
>> > particular S/T-PE and that is straightforward.
>>
>> I agree.
>>
>> >
>> > Note that TTL expiry VCCV is not inband when CW is not used since
>> > intermediate nodes can look beyond the label stack and have different
>> > ECMP behavior.
>>
>> Exactly. Which prompts the question:
>>
>> If the VCCV packet takes a different path than the PW data, can this
>> different path have a different number of hops and have the VCCV expire
>> at a PE different than the target PE?
>>
>> > A draft that addresses this condition by extending TTL
>> > expiry VCCV was submitted at the last IETF
>> > (draft-kini-pwe3-inband-cc-offset).
>>
>> Scanning now through this I-D, I could not find why a 'flow/entropy
>> label' [fat-pw] would not work.
>>
>> Thanks,
>>
>> -- Carlos.
>>
>> >
>> > Thanks
>> _______________________________________________
>> pwe3 mailing list
>> [e-mail]
>> https://www.ietf.org/mailman/listinfo/pwe3
>
>
> _______________________________________________
> pwe3 mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/pwe3
>
>



-- 
Sriganesh Kini (Sri)