ws-sandesha-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Chamikara Jayalath <chamikar...@gmail.com>
Subject Re: Exponential backoff Policy
Date Fri, 11 Nov 2005 10:17:33 GMT
oops. OK :-)
Seems like the best way is to make it plugable.

Thanx,
Chamikara



On 11/11/05, Paul Fremantle <pzfreo@gmail.com> wrote:
>
> The WS-RX committee has just removed exponential backoff from the working
> spec :-)
>
> The committee has added the following wording to the spec (roughly)
>
> "Consequently, implementers are encouraged to utilize adaptive mechanisms
> that dynamically adjust re-transmission time and the back-off intervals that
> are appropriate to the nature of the transports and intermediaries
> envisioned. For the case of TCP/IP transports, a mechanism similar to that
> described as RTTM in RFC 1323 [RTTM] should be considered."
>
> So based on this I would try to base Sandesha's on RFC1323.
>
>
> Paul
>
>
> On 11/11/05, Philipp Reinecke <preineck@informatik.hu-berlin.de> wrote:
> >
> > On Fri, Nov 11, 2005 at 01:44:49PM +0600, Chamikara Jayalath wrote:
> >
> > Hello,
> >
> > > Does anyone exactly know how adjust the retransmission intervals when
> > > exponentionBackoff=true.
> > >
> > > What I have in mind is , if the base retransmission interval is 400
> > ms. the
> > > message retransmission interval should be adjusted as
> > > 400,800,1600,3200,..... etc for successive retransmissions of the same
> > > message. Is this the correct interpretation ???
> >
> > I seem to recall that I asked a similar question a while
> > back, and while I do not have a definitive answer, I tend
> > to agree with your interpretation:
> >
> > (a) Just do a simple binary exponential backoff, i.e.
> > double the interval on every retransmission.
> >
> > (b) Back off for individual messages, not whole sequences.
> > This is because there is no way to shrink the interval with
> > exponential backoff, therefore you might soon end up with a
> > very large interval and all its associated problems, foremost
> > that of not being able to detect loss on time, just because
> > a few messages had to be resent.
> >
> > > Thanx,
> > > Chamikara
> >
> > Regards,
> >
> > Philipp Reinecke
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: sandesha-dev-unsubscribe@ws.apache.org
> > For additional commands, e-mail: sandesha-dev-help@ws.apache.org
> >
> >
>

Mime
View raw message