A thousand apologies, my email client gets mad. So, one more time:

If we have one thread why not to use some time-expired flags?

Say, after sms was submitted, set "mute_submitting_till" to (now +
1.0/throughput), where now = current time in usec

Then if kannel wakes up, it do incoming data processing, and if it
wakes up before (mute_submitting_till < now) == true, he don't send
anything and do gwthread_sleep(now - mute_submitting_till)

Stas.

On 5/13/07, Vincent CHAVANIS <[EMAIL PROTECTED]> wrote:
Yes this is true. We have an old issue here.
We are trying to find a global solution for this but this is not so trivial.
You can use/adapt a patch (for EMI/UCP Smsc) i posted few month ago to the
ML
If i remember well, Stuart Beck has already adapted it for SMPP.
But we get a new issue after using this patch, the Throughput will be
respected
but you will limit the incoming stream because we have only one thread that
do RX and TX

Vincent

----- Original Message -----
From: "Stanislaw Senotrusov" <[EMAIL PROTECTED]>
To: <[email protected]>
Sent: Sunday, May 13, 2007 10:52 AM
Subject: bug in throughput?


> Hello.
>
> My SMPP connection is configured with throughput = 1
>
> Looking at logs I see what kannel trying to do submit_sm more than 1 per
> second.
>
> I have a 1 message per second limit, and carrier SMSC responds to my
> high volume message flow with SMSC returned error code 0x00000008
> (System Error) in response to submit_sm
>
> As I noticed, throughput handled through gwthread_sleep() function
> which is unblocked on incoming data in socket.
>
> So, kannel do submit_sm, and then trying to sleeps for 1 second, but
> that sleep is breaks by incoming data from SMSC.
>
> Am I right or I can't get something?
>
> --
> Stanislaw Senotrusov
>
>





--
Stanislaw Senotrusov

Reply via email to