Which CPD for Outbound?

[PFCCWA]

Hello,
Its been a while since there was a poll about this and I wonder if things have changed.
What CPD are you using for Outbound, is it working effectively?

I have a client who is still using hardware (dialogic) for cpd and want to convince them to change and use software solution (preferably GVP) - is the best route?  I have deployed this on another site and it worked really well in addition to being easy to set up.  Only problem client is not using SIP (t-server with Avaya/h.323 inbound).  I still think dialogic card should be replaced with possibly HMP but think this still needs IP solution.

thanks,

[cavagnaro]

I tried Sangoma and worked very well. More customization which is what I like.
I heard that MediaServer also works fine.

HW detectors have today a lot of restrictions. Not recommened anymore for cellphones detections as there is now on the market many more complex situations than used to be 5 or 10 years ago. HW were designed for land lines, which now are rare to make contact to.

[PFCCWA]

Just so I am not misunderstanding..
Is GVP media server only available as cpd in SIP environment?
Client outbound records are all cell, with landline as secondary option so plan to recommend gvp -if possible or sangoma.
Their inbound lines are converted from ISDN to h.323 but not sure if this qualifies for ip solution.

thanks.

[cavagnaro]

SIP is the way to go. As all those new solutions are based on RTP analysis you don't need HW anymore. Doing HW conversion takes time and resources, things too valuable on CPD you can't just waste them.
Not sure how H323 arrives to the actual CPD...or Genesys at all. You would have to get a MG to do the H323 to SIP or go directly from ISDN PRI to SIP, easier and fastest way. No middlewares. Sangoma offers a HW ready to do conversion too.

[genesysguru]

Good results with Sangoma which improve over time.

[Fra]

Interesting to get to know that some of you have had a positive experience with Sangoma; personally, other users didn't provide the same feedback.

With regards to SIP, I'm not happy with what I have seen so far with regards to the impact on SIP-based outbound diallers; in fact, although CPA takes place mostly on the RTP layer, SIP signalling has a significant role in the call progress detection phase. However:
- SIP is not a *standard* ; that means  hw / sw vendors implement it differently
- service providers configure their equipment as they wish, hence the SIP signalling returned may differ even under the same use case / scenario
- although SIP has per se a significant number of error codes, in reality just a handful of them are used, this leads to providing a less granular range of results
- a cascading inter-protocol translation just contributes to losing accuracy on what happens to a transaction.

What does that mean? That a call to an out of coverage mobile number allocated to service provider X may get a 480, whereas service provider Y may well return a 503 followed by a 'the number you've dialled is not reachable' announcement, both mapped differently by Genesys OCS.
Clearly, dialler business users would get their stats skewed and biased without having an understanding of the underlying root cause.

Fra

p.s. I would be very keen to see the answering machine / live person detection performances of Media Server in a regulated environment (UK OFCOM, US FTC etc): can anyone share the details please?

[cavagnaro]

Hey Fra, those scenarios you describe are mainly part of the integration between media gateway and PSTN provider or media gateway and sip server. Sangoma just is used as a extra layer between sip server and your media gateway. Now, I have used Audiocodes and Alcatel OXE,  haven't had those scenarios and even if I did, I'd work on the MG integration and not Sangoma layer. My thoughts only

[Fra]

Cav,

although the integration between Sangoma or another 'middle-element' between SIP Server and the media gateway is key and influences the signalling, I'm afraid I disagree this time. 
The scenarios I described above are all related to what happens to calls transversing the PSTN and to what telecoms / service providers of the called number return, so [u]upstream[/u] of the media gateway. Of course, configuring Sangoma and or the Media Gateway to re-map the messages may work and help, but still, requires additional effort to level the scenarios.
All I was trying to say is that where with ISDN you'd have consistent results, with SIP that can't be given for granted.

Fra

[cavagnaro]

Ahh of course, I agree with you
PSTN will be always easier to handle, but as premise configuration not always too flexible for new tech. 

[PFCCWA]

Has anyone used Dialogic HMP Software with T-Server?
Ive used with SIP, but never seen it work with T-Server.
I have customer using Avaya AES/Dialogic hardware/T-Server/CPD Server/OCS 8.1.1 but if we were to move away from h/w can we use HMP.
thanks.

[Dionysis]

I've done comparison tests between Dialogic HW CPD and Sangoma.  Media Server uses the same logic as Dialogic HW cards so I wouldn't expect to see any difference at all here.

Sangoma wins, by a long way!  But it does need a little tuning to begin with.

If you break calls up into 3 categories, Business, Home, and Mobile, we got the following results.  This is with 1 list split randomly across the two methods.

Dialogic HW Correct Detection %:
Business: 80-82%
Mobile: 85-88%
Home: 90-92%

Sangoma Correct Detection %:
Business: 92-94%
Mobile: 93-96%
Home: 95-98%

These stats were achieved AFTER tuning the Dialogic settings extensively, over a period of several months and hundreds of thousands of calls. The tuning of the Sangoma setup was done within 2 weeks and required a minor tweak to the confidence level settings and shortening the detection timeout to bring the system within legal boundaries.

I wouldn't use anything else.

[Fra]

[quote author=Dionysis link=topic=8559.msg37766#msg37766 date=1414451804]
I've done comparison tests between Dialogic HW CPD and Sangoma.  Media Server uses the same logic as Dialogic HW cards so I wouldn't expect to see any difference at all here.

Sangoma wins, by a long way!  But it does need a little tuning to begin with.

If you break calls up into 3 categories, Business, Home, and Mobile, we got the following results.  This is with 1 list split randomly across the two methods.

Dialogic HW Correct Detection %:
Business: 80-82%
Mobile: 85-88%
Home: 90-92%

Sangoma Correct Detection %:
Business: 92-94%
Mobile: 93-96%
Home: 95-98%

[/quote]

Dionysis,

thanks for posting this. I do have some questions though
[list]
[li]what was the time you reserved to the call progress detection timeout?[/li]
[li]does Sangoma have the ability of firing off that timeout on the signalling event 'connect' or at the start of voice energy?[/li]
[li]do the results you've published here relate to detecting live voice or answering machines?[/li]
[/list]
Thanks

Fra

[cavagnaro]

On question #2 you can do both

[Dionysis]

Sangoma doesn't use "voice energy", it matches wave patterns.  Which is what sets it apart from the Dialogic and Media Server method.

The timeout we used was 4 seconds, we have a legal requirement (in Australia) that calls cannot take longer than 5 seconds in total from answer to agent, including CPD and possible queue time.

[Fra]

[quote author=Dionysis link=topic=8559.msg37807#msg37807 date=1414735052]
Sangoma doesn't use "voice energy", it matches wave patterns.  Which is what sets it apart from the Dialogic and Media Server method.
[/quote]
Ok, what I was interested in was whether Sangoma can be set to start the timeout when it detects speech or not, which Cav said it does.
[quote author=Dionysis link=topic=8559.msg37807#msg37807 date=1414735052]
The timeout we used was 4 seconds, we have a legal requirement (in Australia) that calls cannot take longer than 5 seconds in total from answer to agent, including CPD and possible queue time.
[/quote]
Lucky you Ozzies! Regulations mandate a 2 secs timeout here in the UK, hence I'm reluctant to be too positive about the potential results here..

Fra