RE: [PWE3] PWE3 protection draft

[Sasha Vainshtein <[e-mail]>]

Ping, 
I have several questions about your draft. 

Q1. What exactly is the scope of the "Protection PW" technology? 
    I read the text as if it equally applies to SS- and MS-PWs. However, 
    in the case of a single-segment (SS) PW, a "protection tunnel" 
    between the PEs seems to be enough, you do not need a separate 
    PW label. In the case of multi-segment (MS) PWs, you may need 
    a protection PW if you want to safeguard against the failure of 
    one of the S-PEs. Is this understanding of scope (i.e., only MS-PWs 
    that require protection against S-PE failures) correct?

Note: Specific requirements (e.g., downtime etc.) seem to be lacking
      as well. Without understanding these requirements, it is
      difficult to analyze the proposed solution.

Note: PW-level protection vs. tunnel-level protection comparison
      looks similar to SNCP (a.k.a. ULSR) vs. MSP-Ring (a.k.a. BLSR)
      SONET/SDH protection schemes. This analogy may be useful for
      analysis of the proposed scheme. 

Q2. IMO PW protection needs to be bi-directional, i.e. if one U-PE 
    switches from the "main" PW to the "protection" one, its peer 
    should do the same. The draft does not say how this is achieved 
    (indeed, IMHO it does not specify any protection switching
    procedures at all). 

Note: The draft does not mention how the PW failure (presumably 
      triggering the switch) is detected. It is also not clear if
      forced switch can be used, e.g., ahead of planned shutdown
      of one of the S-PEs etc.

Q3. The Generlized PW ID FEC comprises PW Type, AGI, SAII and TAII 
    (see draft-ietf-pwe3-control-protocol-17.txt, Section 5.3.2). 
    Your draft says that these must be the same for the Main and 
    Protection PWs, implying that 2 labels must be allocated for the 
    same Generalized PW ID FEC both by the U-PEs and S-PEs. 
    This looks to me like a somewhat unusual arrangement. Did you. 
    consider using two different FECs and signaling the 
    "main/protection" relationship between these two FECs instead? 
    This would restrict the processing of additional information to 
    U-PEs only and separate signaling and routing of the "main" and 
    "protection" PWs. One way to do so would be establishing the 
    "main" PW first, learning its S-PEs and then setting up the 
    "protection" PW that avoids using the same S-PEs. 

Hopefully these notes will be useful. 

Regards, 
              Sasha 

 

-----Original Message-----
From: Ping Pan
To: [e-mail]
Sent: 3/18/2006 4:03 AM
Subject: [PWE3] PWE3 protection draft

Hi,
 
Due to some problems with headers, this draft was not posted to after
three weeks.
 
http://www1.cs.columbia.edu/~pingpan/papers/draft-pan-pwe3-protection-02
.txt
<http://www1.cs.columbia.edu/~pingpan/papers/draft-pan-pwe3-protection-0
2.txt> 
 
Would like to present this next week.
 
Many thanks!
 
- Ping
 <<ATT328895.txt>> 

_______________________________________________
pwe3 mailing list
[e-mail]
https://www1.ietf.org/mailman/listinfo/pwe3