Re: [M3UA] AS state machine sharing between SGPs

Tolga,

This excerpt should make it clear where the statements in the RFC came from.

It is good to refer to the mail archive when memory fades.

--brian

Sidebottom, Greg [BCNR:7M01:EXCH] wrote:                       (Tue, 15 May 2001 14:31:25)
> 
>    [Greg]  I think it is fair to conclude that having multiple SGPs in an
>    SG  does  not  in  itself  buy  any  extra  functionality  in terms of
>    redundancy  over  the  multiple  SG model.  I think we've been bending
>    over   backwards   in  the  protocol  description  to  accommodate  an
>    implementation  option  (the  "distributed  multi-site  SG"  described
>    above).    So  where  does  this implementation option make sense? - I
>    think the answer is for the case where the signalling between the PSTN
>    and  IP  networks is to go through a single SG with a single SS7 Point
>    code but with site/processor redundancy for the SG.  Paired SGs won' t
>    do  because  two  Point Codes would be required.   I don't really have
>    problem with supporting this case.
> 
>    I  think  we  can  change a few things in the spec to support this but
>    remove some of the confusion that it brings:
> 
>    1)  Simplify the protocol description so it is described as between an
>    SGP  and ASP only.  That is, SCTP associations are always between SGPs
>    and  ASPs.   Today the spec goes back and forth between the SGP and SG
>    terms and adds qualifying text in various places.
> 
>    2)  The  SG  becomes  explicitly  a  virtual  thing  not unlike an AS,
>    containing  one  or more SGPs.  The spec today seems to shade back and
>    forth on whether an SG is a physical thing or not.  Of course if an SG
>    is implemented using a single SGP, then that is fine  2;^)
> 
>    So  a  paired  SG  architecture  is  now  two  SGPs on different hosts
>    represented  to  the  SS7  network  with  a  Point  Code per host.  No
>    coordination  of  MTP3,  SPMC  or  AS/ASP  states  between the SGPs is
>    required.   A  single  distributed  SG  is  the same except that it is
>    represented  to  the  SS7 network with a single SS7 Point Code for the
>    set  of  SGPs.   The  SGPs  SHOULD support MTP3, SPMC and AS/ASP state
>    sharing  and  re-routing  of  traffic  between  the two SGPs.  How the
>    inter-SGP  protocol accomplishes this is not described in the protocol
>    spec.
> 
>    3) The ASPs do not need to know the grouping of SGPs in virtual SGs so
>    they  treat  each  SGP as an independent route out to the SS7 network.
>    But an ASP MAY choose to maintain such knowledge.   The SSNM messaging
>    from  the  SGPs to the ASP is therefore processed independently.  Note
>    that  for  Operators  that choose not to combine SGPs into virtual SGs
>    (i.e.,  use  paired  physical  SGs),  then  nothing has really changed
>    here.    ASPs  that  choose  to  keep  knowledge of SGP groupings into
>    virtual SGs MAY interpret SSNM messaging from one of the SGPs so as to
>    apply  to  all the SGPs in the virtual SG but should only do so if the
>    SGPs coordinate their SSNM messaging to the ASPs.
> 
>    Hope this helps,
> 
>       Greg Sidebottom

-- 
Brian F. G. Bidulock
[email protected]
http://www.openss7.org/