[Asterisk-Dev] Who is developing Shadydial?

C. Maj cmaj-SPAM at freedomcorpse.com
Thu Apr 15 12:50:40 MST 2004


On Thu, 15 Apr 2004, Richard Airlie waxed:

> On Wed, Apr 14, 2004 at 09:47:46AM -0400, C. Maj wrote:
> > On Wed, 14 Apr 2004, Richard Airlie waxed:
> > 
> > > I'm interested in finding out a bit more information about how well shadydial
> > > performs, what sort of real world usage it's seen, etc. Can anyone shed any
> > > light?
> > 
> > I pulled out a showstopper bug almost 2 months ago.  Since
> > then, I've been running it in testing for hours and hours
> > at a time, about 3-10 agents.  The box is connected over IAX
> > to the production server so I can still beat on it, reboot
> > often, etc.
> 
> Is that 3-10 real agents doing real work on it (making live calls,
> recording call outcomes). Are they connected to Asterisk over the PSTN
> or VOIP? (or is that of no consequence?)

Yes.  They look real to me, at least.  They use analog
phones, connected to a channel bank, to a T400 on the
production * box, then over IAX to the * box running
shadydial.  Calls originate from this testing box, over IAX,
out the production box via a T1.

> How is the predictive dialing algorithm performing? Does it keep the
> agents quite busy with calls? What sort of abandon rate are you seeing,
> and is it possible to set a target abandon rate?

It's alpha software.  The algorithm currently relies on
several figures as collected over the previous hour, like
average call length and answer percentage.  These are
refreshed every minute and also factored with number of
agents logged in.  There's a little more to it, but that's
the meat of any predictive dialing algorithm.  At least as
far as I can tell.

> > Still a long way to 1.0, tho.  But I think the next release
> > will be a beta instead of alpha (or VHS).
> 
> Sounds great. What are the most 'alpha' areas of Shadydial..
> what parts need the most work?

Well I'm trying to make this as ghetto as possible.  Meaning
no computer agent stations, just a regular phone line.  That
means call outcomes are recorded in the agent hang up
function, so it blocks terribly.  That's why it doesn't
scale.  A fix is on the todo list, but it's not really
necessary for my purposes for it to scale any larger than 10
agents.

If, however, you weren't trying to go ghetto like that, you
could rip out the result capturing, give everyone a
computer, and send an URL with a phone number to pull data
to a web browser on the agents monitor.  I think somebody
may already have done this.

> Hopefully I'll have a test box for Asterisk quite soon and will
> be able to spend some time looking at and playing with shadydial.

Good luck, keep me posted.

--Chris


-- 
Chris Maj, Rochester
cmaj_at_freedomcorpse_dot_com
Pronunciation Guide: Maj == May



More information about the asterisk-dev mailing list