Re: sctp retrans count

"Brian F. G. Bidulock" <[email protected]>
Newsgroups gmane.ietf.tsvwg,gmane.ietf.sigtran
Organization http://www.openss7.org/
Message-ID <[email protected]>
devayya,

M2PA T7 has an impact on MTP3 running above it.  A stock MTP Level 3
cannot accept T7 greater than 2 seconds without serious impact on MTP's
ability to changeover links correctly.  For example, Q.704 T1 and T2 are
1 to 2 seconds.  MTP Level 3 expects that after 2 seconds (or so) all of
the messages for a given link have cleared or the link has failed.
Increasing M2PA T7 to greater than 2 seconds breaks this condition and
would allow old messages to persist in the network.  This could increase
MTP Level 3's duplicated message probability above what is an acceptable
quality of service for MTP management or MTP Users.  Therefore, M2PA T7
should not be increased above 2 seconds without giving consideration for
MTP Level 3 timers and the quality of service expected by MTP Users.

I have seen adding delays into links in the same manner as increasing T7
cause cascading failure and full isolation between STPs.  When
increasing T7, buffer thresholds should be reduced drastically.  For
example, you say that you are running 5000 msus/second.  If you are
running 12-bit sequence numbers (ala Q.703 Annex A) only 4096 MSUs can
be oustanding and more than 1 second ack delay will cause the RTB to
fill and TB occupancy to increase.  If you are running 24-bit sequence
numbers (ala Q.2210), the RTB will not be force to fill due to the
sequence number window, however, say 4 seconds could cause 20000 MSUs to
be waiting for retransmission.  Performing buffer collection and
retransmission on an alternate link of this many MSUs could cause
congestive failure of the alternate link which can cascade to the
failure of all links between two signalling points.  It is better to
keep T7 low and fail links before buffers fill.

M2PA will likely not reliably run with a stock MTP Level 3 over anything
but an engineered newtork.  Care must be taken that M2PA provides the
same link reliability to MTP Level 3 as it expects from other links and
should fail the link otherwise.

--brian

devayya wrote:                                 (Fri, 19 May 2006 16:53:14)
> 
>    Brian,
>    Thanks for that calculations and update.
>    Does  this  mean  that  suppose  we  decide to change our default SCTP
>    parameters  to the one recommended in rfc 2960, Should we increase the
>    m2pa T7 (delay ack timer) timer value to a higher value?
>    Presently  the  m2pa  T7 timer value of 2 seconds is the maximum value
>    that  I ahd mentioned that can be provisioned in our setup. By default
>    T7 timer is only 1 second.
>    Thanks and regards,
>    devayya

-- 
Brian F. G. Bidulock
[email protected]
http://www.openss7.org/
lmpx.com only provides a reader for public news (NNTP) servers. It is not affiliated with the servers or forums shown here and is not responsible for the content of articles, which is written by their respective authors.