Re: Fwd: Re: SE style in optical neyworks

[Suresh Katukam <skatukam@cisco.com>]

Hi Zhi,

Please see my comments in-line.

Here is an example:

>>>          1+1                        1+1
>>>A ----------------B --------- C ------------D
>>>                      \          /
>>>                        \       /
>>>                           E
>>>
>>>
>>>A - B & C - D 1+1 line protected
>>>B - C, C - E, B - E are all unprotected links.
>So this network you drew assumes a single point of interconnect at B and 
>C...that's fine

Yes.



>>>Say, one wants to create a protected LSP from A to D.
>>>then, you would create one primary LSP from A to D via
>>>A - B - C - D, and then you would create another LSP
>>>from A to D using SE style ( to indicate that this is an
>>>alternate path for same Tunnel ) via A - B - E - C - D.
>>>This way B - C is protected by B - E - C.
>By creating a new LSP A-B-E-C-D, this circuit is completely separate from 
>the original A-B-C-D circuit. There is no sharing at all. It is just 
>incidental that they carry the same traffic (i.e., you just happen to have 
>set up the A-B-E-C-D to carry the same traffic as in A-B-C-D). Again this 
>is not merging (as represented by SE), unless you are re-defining what SE 
>means...

As others pointed out, SE may not be a right terminology to be used for 
this. There is a merging happening at node B since
B uses line APS to find out active link and selects and bridges traffic to 
E and C. Certainly it is not like packet merging, but
the traffic terminates at B from both lines and only one link traffic is 
sent out to E and C. So I am not sure
what do call this? But looks like SE may not be a right term. As you all 
know that, on 1+1 line you make only one
reservation for a LSP, even if LSP is protected. So if one wants to create 
a protected LSP, only the first LSP
makes reservation on 1+1 link and second LSP does not make any additional 
reservation.




>>>B - C - E is not configured as a UPSR. All unprotected links.
>>>Virtual UPSR can be created such a way that B has a bridge
>>>and C has a selector. Ofcourse, for bidirectional LSP, you need
>>>the same thing in the opposite direction too.
>Yes, just as the two LSPs making up the 1+1 from A-B are both unprotected 
>LSPs, but taken together provide the 1+1 service.

Nope. Both LSPs are protected on A-B link. Even if you do not create second 
LSP, the first is protected by default on
A-B due to 1+1 line APS.

>  As such, if B-E-C is carrying the same traffic as B-C *all the time*, 
> then it's still 1+1, but more specifically, it is a 1+1 SNC (sub-network 
> connection). 1+1 doesn't necessary only apply to 1+1 path (when you say 
> path, I think you mean trail? -- G.805 terminology).
>
>However, if B-E-C is not carrying the same traffic, but may be used to 
>support *extra traffic*, then this is a 1:1 type protection.
>
>And if B-E-C LSP is actually supporting multiple LSPs going over B-C, then 
>it's 1:N protection. However, note that even in this case it is still NOT 
>SE, since the B-E-C LSP at any given time is only ever carrying one single 
>traffic.

I agree with this..

Thanks,
Suresh


>Hope this clarifies.
>
>Zhi
>
>>>
>>>Now, what do you call this kind of protection? Protected LSP?
>>>But not 1+1 path protected  - since it is not end-to-end path
>>>protected.
>>>
>>>To clarify again, this is when you would use SE style.
>>>
>>>Thanks
>>>Suresh
>>>
>>>
>>>At 01:17 PM 4/12/2002 -0400, Thomas D. Nadeau wrote:
>>>
>>>
>>>>>This can be used to set up Protected circuits which may contain
>>>>>1+1 lines and 1+1 path protected segments. So on 1+1 line
>>>>>protected segments, you Share the bandwidth among primary
>>>>>and alternate paths.
>>>>
>>>>
>>>>         You do not share the bandwidth across 1+1 circuits. You
>>>>double-book the bandwidth because the same packets are
>>>>sent twice: once over each LSP.  You only share bandwidth
>>>>with 1:N.
>>>>
>>>>         --Tom
>>>>
>>>>
>>>>>Setting up protected circuits is not considered
>>>>>in detail so far. Hopefully, P&R design team will consider this..
>>>>>
>>>>>Thanks,
>>>>>Suresh
>>>>>
>>>>>At 04:32 PM 4/12/2002 +0530, Khuzema Pithewan wrote:
>>>>>
>>>>>>Hi,
>>>>>>
>>>>>>How does SE style of RSVP signalling fits in the optical nature of 
>>>>>>network
>>>>>>i.e. in wavelength, TDM switching etc.
>>>>>>
>>>>>>  In other words, How two lsp can share resource in optical networks??
>>>>>>
>>>>>>Regards,
>>>>>>Khuzema.
>>>>
>>>>
>>>>
>>>>------------------------------------------------------------------------
>>>>Mathematics is the supreme nostalgia of our time.
>>
>