Fw: [Users] TM : retransmission timers

Mike Williams mike at mikebwilliams.com
Wed Nov 8 18:38:40 CET 2006


Kim,

Here are a few features of the top of my head that I can think of, both fetch 
support and new a DB mode for usrloc. 

The new db mode allows me to have two OpenSER servers load balanced using just 
DNS SRV records and a mysql backend because any SIP transmission can go to 
either server, and any server can fail without effecting anything else.

From the openser home page:

Accounting optimization:

"2006-09-25 - Accounting module (acc) has been redesigned to be more flexible 
and faster. 

 The module suffered major changes in the internals, but kept easy interface 
to the configuration script. Lot of useless fields were removed from the 
default list, they can be recorded on demand via extra accounting facility. 

 Many other optimization were introduced to enable modularity and faster 
storage. Also, multi-leg accounting was enhanced to allow definition of as 
many fields to be recorded as needed."

This describes some major memory tunings as well better support for huge 
numbers of users:

Database fetch support    2006-07-14: 
 New introduced fetch support for database drivers enables management of huge 
number of location entries (>120 000) with no need of OpenSER compilation 
tunings. 

 Just few days after the last major release , new appealing feature was 
introduced in the development version. 

 Until now, when dealing with large number of subscribers, OpenSER/SER 
required tunings and recompilation to resize the private memory large enough 
so that all location records from database could be loaded. It was quite 
annoying because this private memory was used only once, at startup, 
afterwards most of it was not used at all by the SIP server, but it was 
allocated and no other application could use it. 

 The new features loads chunks of records, the size of the chunk being 
configurable and specific for the available private memory. The issue is now 
solved in OpenSER, allowing straightforward deployments for carrier grade 
platforms. This adds more value to the features introduced with the latest 
release that enabled geographic distribution/failover and load balancing.

If I think of some other good stuff, I'll let you know.

Perhaps someone should put all these comparisons into a wiki page so that we 
can build an accurate picture of the advantages of each product.

Mike Williams

On Wednesday 08 November 2006 11:46, Kim Il wrote:
> Mike,
>
> thanks a lot for your insight and the time you have taken to answer my
> question. I surely did not want to be unafir to anyone. I can not comment
> on the why the SER people are having only few releases compared to openser
> but I guess they have a good reason for that (I would guess it has
> something to do  with thoroughness, testing ..., but I am here just
> guessing). For me and I guess many other people who depend on SER/openser
> for their business the more interesting question is what to use in the long
> term, which camp is the more innovative, is actually driving the change and
> is commiting itself to the really difficult tasks. Your statement that
> changes in SER have been  in response to openser is surely a very good
> starting point -and if this is true then there is no reason to shift back
> to SER. Do you, or maybe other people, have more examples -besides the web
> page- for this. This will really help me to better judge the situation.
> Looking at the release notes of both openser and SER I am still of the
> opinion that compared to the new features of SER, most of the features that
> were added to openser in the last year are of cosmetic nature (hardly any
> touches the tough issues of security, performance and reliability).
>
> Yours sincerely
>
> Kim
>
> Mike Williams <mike at mikebwilliams.com> wrote: Kim,
>
> I don't think that's really a fair assessment of the situation. It seems to
> me, and others much more knowledgeable please comment on this, that OpenSER
> was forked because SER was left to stagnate, and because of a large number
> of feature patches that were just left to sit. The development cycles
> became too long, and it was unclear what the plan was.
>
> Looking back on the progress of OpenSER, one can see that the team didn't
> just take those patches and merge them, and pretend that they have a new
> product, but have instead continually developed the code base. The always
> have a roadmap of the next release, and an estimated timeline for
> completing it. A lot of important features have been added.
>
> Likewise, OpenSER seems to be using a different development philosophy. The
> OpenSER team releases .1 increment releases with new, useful, and stable
> features fairly often instead of waiting years. Since I've been using
> OpenSER, I've seen 3 releases. SER has put out 1 in that same time period,
> and honestly, I don't see the same amount of features really being added by
> SER. If anyone can compare the two in their present STABLE forms, I would
> really like to hear about it.
>
> In addition, it seems many of the changes to SER have been in response to
> OpenSER. Iptel/SER had the same website for years, with little information
> about what was actually happening. If you check the OpenSER website, they
> are always giving useful information and news to the users and community
> about going forward. Just in the last few months has Iptel/SER actually
> changed, no doubt partly due to how good OpenSER looked in comparison.
>
> Mike Williams
>
> On Wednesday 08 November 2006 04:06, Kim Il wrote:
> > thanks Rao for bringing this up. Actually our company has moved to
> > openser around 6 months ago after reading the rumor spread around on the
> > openser lists that SER is no longer being maintained. Looking now at the
> > new SER I must confess that we are more than impressed  about the new
> > features and substantial changes  to SER. It seems that, unlike openser,
> > the guys
> > behind SER spent the time not on cosmetic and superficial changes but on
> > real improvements. I assume this difference in working style comes from
> > the fact that openser is lead by a company that is capitalizing the
> > open-source spirit to satisfy the day-to-day needs of it customers
> > whereas SER is being maintained by guys who have a long term vision of
> > things. While it will surely cost us some time and effort for us the
> > decision is already clear that unless openser integrates the SER
> > improvements we will go back to SER.
> >
> > Bye
> >
> > Kil Il
> >
> > Rao Ramaratnamma  wrote: sorry for reposting -- I
> > think this question belongs to both mailing list. I am really stuck with
> > this.
> >
> > rr
> >
> > ----- Forwarded Message ----
> > From: Rao Ramaratnamma
> > To: Christian Schlatter ; users at openser.org
> > Sent: Tuesday, November 7, 2006 11:15:27 PM
> > Subject: Re: [Users] TM : retransmission timers
> >
> > the ser ottendorf announcement does mention improved timers. Cannot
> > openser include this feature too and cannot I merge ser  with openser for
> > good timers? I am still trying to understand the difference between ser
> > and openser but standart compliance seems to be very important matter!
> >
> > Cannot people provide me with some hints? I am sure that I am not the
> > only who is asking the difference between ser and openser. ser
> > documentation does not appear uptodate, but the software as sannounced
> > appears impressive. I have already asked this question but did not
> > receive any answer.
> >
> > thank you in advance!
> >
> > rr
> >
> > ----- Original Message  ----
> > From: Christian Schlatter
> > To: users at openser.org
> > Sent: Tuesday, November 7, 2006 10:52:56 PM
> > Subject: Re: [Users] TM : retransmission timers
> >
> > Greg Fausak wrote:
> > > Hello,
> > >
> > > I believe this is a well known bug.
> > > Granularity of timers is 1 second.  So, if you sign up for a timer to
> > > be fired in 1 second it will happen anywhere between 0 seconds and 1
> > > second.
> > > 2 seconds will happen between 1 and 2 seconds.  I usually set up my
> > > timers to be 2, 2, 4, 8.  There are VOIP providers that are pretty
> > > sticky about
> > > the first 500ms.  If you are using one of them you're out of luck.
> >
> > Yes, there is a timer process that wakes up every second to perform
> > retransmissions. I was actually quite surprised that OpenSER, which is
> > known to be very standards compliant, does not follow the RFC 3261
> > retransmission timeouts. On  the other hand, the RFC 3261 timeout values
> > are just suggestions and standards compliant SIP UA must accept shorter
> > timeouts. Still it would be nice if OpenSER would support sub second
> > timers, this would allow for shorter fail-over times.
> >
> > Christian
> >
> > > I believe SER has made timer changes to support more exact timer
> > > intervals.  They are a completely different camp, with a different
> > > feature set (although they share the same roots).
> > >
> > > -g
> > >
> > > On 11/7/06, Jean-François SMIGIELSKI  wrote:
> > >> Hello,
> > >>
> > >> I made strange observations about the intervals between
> > >> retransmissions with the TM module.
> > >> In my experiments,  I used the default parameters for the TM module
> > >> timers, and I  sent an INVITE that cannot receive answers (it has a
> > >> well  known R-URI  pattern that is forwarded to a place and port that
> > >> nobody listen).
> > >>
> > >> When reading RFC3261, I expected to see intervals between
> > >> retransmissions of |500ms|1s|2s|4s|8s|16s|. 7 transmissions, during
> > >> 32s.
> > >>
> > >> But with OpenSER, (I have tested with the debian package 1.1.0-5 on a
> > >> debian etch, and the cvs sources for 1.1.0 or 1.0.1compiled by
> > >> myself), I can see intervals like <500ms, 2s, 4s, 4s,4s, ... until 26s
> > >> are spent (9 sendings). The first interval is sometomes very short
> > >> (40ms).
> > >>
> > >> Altough I like the sequence of 4s separated transmissions, I do not
> > >> know why the first interval is so short, and why there is no sending
> > >> after 1s.
> > >>
> > >> Did anybody observed such behaviours? Are they normal?
> > >>
> > >> Thanks in advance!
> > >>
> > >> JF  Smigielski.
> > >>
> > >>
> > >> ______________________________________________________________________
> > >>__ iBELGIQUE, exprimez-vous !
> > >> http://web.ibelgique.com/
> > >>
> > >> _______________________________________________
> > >> Users mailing list
> > >> Users at openser.org
> > >> http://openser.org/cgi-bin/mailman/listinfo/users
> >
> > _______________________________________________
> > Users mailing list
> > Users at openser.org
> > http://openser.org/cgi-bin/mailman/listinfo/users
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> > _______________________________________________
> > Users mailing list
> > Users at openser.org
> > http://openser.org/cgi-bin/mailman/listinfo/users
> >
> >
> >
> > ---------------------------------
> > Everyone is raving about the  all-new Yahoo! Mail.
>
> _______________________________________________
> Users mailing list
> Users at openser.org
> http://openser.org/cgi-bin/mailman/listinfo/users
>
>
>
> ---------------------------------
> Sponsored Link
>
> Get an Online or Campus degree - Associate's, Bachelor's, or Master's - in
> less than one year.




More information about the sr-users mailing list