Re: M2PA (SIGTRAN) and multi-homing

Robby,

Robby Pocornie wrote:                                               (Wed, 06 Dec 2006 15:12:50)
> 
>    Hi Brian,
>    Hi All,
> 
>    Is it useful to have SCTP multihoming on M2PA links between Signalling
>    Transfer Points/Gateway (STPG)?
> 
>    I  see  in  the  IETF RFC that multihoming on M2PA has been taken into
>    account as a possibility.
>    Is this possibility advisible, without being harmful for the Network ?

SCTP multihoming is not harmful to the network (otherwise it would not have
been included in SCTP).
> 
>    I also read that It is possible that transmission of M2PA messages MAY
>    be  delayed  by  SCTP  slow  start under certain conditions, like Path
>    failure (multi-homing) if SCTP switches to a new path.
>    (<draft-ietf-sigtran-m2pa-09.txt >)

Yes, they may be delayed.  But M2PA runs its own ack timer to ensure that
the link is still viable.

> 
>    I'm  now  in doubt,  because  a supplier  of STPG's told me that often
>    customers  do  not  use  multi-homing  for  M2PA  associations because
>    redundancy is already provided at MTP Layer 3 (Changeover procedures).

That is an engineering decision.  If there exists sufficient link redundancy
at MTP3, multihoming might be unecessary.

> 
>    The   supplier   claim   that   M2PA  multi-homing can  actually  make
>    configuration  more  complicated  because  of  timers for RTT (at SCTP
>    level) and COO, CBO at MTP layer 3.

It is true in practice that if SCTP protocol paramters are not set carefully
that they can interact negatively with MTP3.  This is one reason why we
recommended only 1 M2PA link between hosts, because if multiple M2PA links
existed between hosts and all fail at the same time, MTP3 could thrash
between links.  MTP3 pretty much assumes that links fail independently and
is designed to accomodate only a single link fialure in a network.  In
regular SS7, multiple link failures are extremely rare.

> 
>    Do you have any argumentation or practical explanation to this claim?
>    Who  can  advise  me why  to  Implement  Multihoming  OR NOT Implement
>    Multihoming in a Sigtran M2PA Network connection?

Unfortunately, there is no golden rule.  These are operational decisions that
must be made with full knowledge of the topologies involved.  If you have
little operational experience of your own upon which to rely, it might be a 
good idea to rely partly upon the operational experience of your vendor.

--brian


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