Cvp/Ucce design query

Unanswered Question
Jun 15th, 2010
User Badges:

HI,


Site1: (Florida)

Voice gw, CM Pub and Sub1, agents, CVP1, RoggerA, VRU PG, Agent PG. Agent ipphones register with Florida/Virginia CM cluster.


Site 2: (Virginia)

CM Sub2, CVP2,  Rogger B, VRU PG, Agent PG


Site 3: (UK)

Separate CM cluster, CVP, Agent PG, VRU PG (These PGs talk to Rogger A and Rogger B of Site1/2).


Site 4: (India)

Voice gw, Agents. Agents register with UK CM cluster.


Setup running on Comprehensive mode.


This is the same setup what rafi.imran has traded posts on.There are few other questions on which we where working so i am opening a separate thread for this.


Questions.


1 In a case when my CVP 1(FL) is down and wan links are down at the same time which makes CVP 2(VA) of no use can i transfer calls from FL Gw to CVP in UK(Primary) and play some outage message and transfer the call to voice mail in UK.

This same scenario can also trigger in a case when my ports in FL CVP are full and i want the calls to be send to UK CVP.Can i use CVP Primary (UK) for servicing calls from US(FL) and IND(Hyd) if needed toghether.Should A2Q object this?


2 Assuming i can use the above condition and transfer the calls to CVP (UK)(PG2) from VG (FL) )and play the self service and then can i have the calls transfered to CCM (FL) (PG1) which sits on a diffrent PG box to have an agent transfer.


If yes then i could utilize my CVP in UK in a better way.


Thanks


Ganesh R

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
Loading.

In a case when my CVP 1(FL) is down and wan links are down at the same time which makes CVP 2(VA) of no use can i transfer calls from FL Gw to CVP in UK(Primary) and play some outage message and transfer the call to voice mail in UK.


The WAN is down. How can you do that? Do you want your survivability to do a take back and transfer on the carrier then an International call to the UK? Man, who's paying this bill? Just play a message - it's not critical.


Florida has all the goods - what can go wrong there?


If CVP1 box fails, you will be sending calls to CVP2 thanks to your proxy server. Oh, wait a second. You don't have a proxy server - what was the method? SRV off the gateway? (I don't like the absence of CUPS).


Anyway, through some magic you get the calls to CVP2 and all is well.


If PG1A box fails, you have a redundant PG - no problems.


Ports will not be full if you have sized it correctly. How many T1s in FL? How many CVP VXML licenced ports? Are you using Audium apps or just microapps? Do you load balance across the VXML servers for Audium?


None of the scenarios you describe make much sense. You NEVER want to treat a call on the US gateway under the control of CVP in the UK. Your FL site is not an issue.


Do trunks come into VA? No agents there, correct?


UK is more of a problem. What happens when the PG-CC link is broken in the UK?


What happens in India? Are there trunks on that gwy? If so, that's the biggest weak spot because of the distance to the CVP (in the UK) which is remote from the CC.


We have asked before - where are the trunks? Who and how many?


Regards,

Geoff

ganeshraghuram Tue, 06/15/2010 - 15:24
User Badges:

The WAN is down. How can you do that? Do you want your survivability to do a take back and transfer on the carrier then an International call to the UK? Man, who's paying this bill? Just play a message - it's not critical.


Geoff if i  have 3 dial-peer with diffrent preference 1 to FL(cvp),2 to VA(cvp) and 3 to UK(cvp).The 3rd will trigger in a case if FL(cvp) is down and WAN link between FL and VA are down.

I would be using the CVP UK over the wan between US and UK only in case CVP Active FL or ICM Side A FL is down and WAN links are down at the same time.Chances of this happening one in million but can we tackle this if it happens.



Do trunks come into VA? No agents there, correct?


No trunks no agents in VA.


UK is more of a problem. What happens when the PG-CC link is broken in the UK?


Fails over to CC in VA


What happens in India? Are there trunks on that gwy? If so, that's the biggest weak spot because of the distance to the CVP (in the UK) which is remote from the CC.


Yes one gw and a single trunk terminated in India.CVP servers are in UK for it.The latency between HYD---UK is 100 msec and between UK----US is 70 msec rtt.


We have asked before - where are the trunks? Who and how many?


2 GW in FL.1 T1 each on each GW.

1 GW in UK.1 E1 on single GW.

CVP servers are in UK for it.The latency between HYD---UK is 100 msec and between UK----US is 70 msec rtt.


UK to US 70ms round trip? I don't believe it - who is the provider?  Transmission through fibre optics has a limit of about 60% light speed, so about 200x10^6 m/s. That's 200km per millisecond. US to UK is roughly 7000km great circle, but the undersea cables don't follow a great circle. So add about 30%. Say 9100km. That should take about 45ms. Double this for RT, so 90s. Add delays in the routers at each end, and you are more like 100ms. By the time you add the latency in your own network,I'd work on 120ms.


India to the UK 100ms round trip - man, I don't believe that either.  More like 180-200ms.


I asked: "UK is more of a problem. What happens when the PG-CC link is broken in the UK?"


Fails over to CC in VA


How? The network is down between the CC and the PG. Do you have another provider path to VA? Are you carrying the voice on-net?


Don't worry about FL - don't try to deal with double failures. UK needs survivability. India needs more than that.


Regards,

Geoff

ganeshraghuram Mon, 06/21/2010 - 14:49
User Badges:

By the time you add the latency in your own network,I'd work on 120ms.



Geoff the RTT is less than 100 for both FL and VA from UK Servers.PFA.For HYD to UK too the responses are less than 100 ms as commited by the customer.Hyd(IND) systems are not ready yet.I guess the customer has used some internal ip's for Hyd to ping.




Fails over to CC in VA


How? The network is down between the CC and the PG. Do you have another provider path to VA? Are you carrying the voice on-net?


Geoff from US(FL) to UK and US(VA) to UK have an MPLS links backed up with internet links over VPN.In case all the CC(FL)---PG(UK) links are down the failover can happen to PG(UK)---CC(VA) which passes through diffrent gateways.


Don't worry about FL - don't try to deal with double failures.


Ok Geoff forgetting about double failures here.But here is one scenario which can be usefull.US and IND works on diffrent non-overlapping shift to cover 24/7.Now CVP(FL) has 31 ports and CVP(VA) with 31 ports.My CVP Servers in UK too has 31 ports each.Now customer adds on few more trunks in FL and wants to use the UK (CVP) for additional ports/failover ports from US(does not want to buy new ports) as they are not serviced from IND then due to non-working hours in INDIA.Can he use the ports in CVP (UK) and have the calls transfered back to Agent's in FL?



India needs more than that.


Geoff are you reffering to the UK failure completely here as if that happens then HYD is down too as CCM cluster too is in UK.Customer is aware of the risk and has agreed to go with it.


The customer has taken services from Reliance who have Point of presence in many of these locations.So in case of routing the calls from US to Hyd or vice versa in case of a site down emergency they will do so as they are not going to be charged international calls services but it is a long distance calls.

Attachment: 

I don't care what the customer tells you. Look at the physics.


US to UK over MPLS with a backup circuit being a VPN over the open Internet is quite common. I would not put CVP over it though. You admit this is a million to one shot. If you design it in, you need to test it and support it - for something that may never happen. Just play an error message.


You already told us once that this was not a mission critical call center - hence the absence of  that second subcriber. Don't overbuild it.


I hear what you are saying about "follow the sun". Dealing with CVP over long latencies requires a bit of fancy footwork with "fetch audio" on the gateway and you haev to do this anyway - because those trunks in India on the gateway there are fetching their VXML from London. So you need to get this right.


The problem in the US is you don't need these tricks - the calls are normally getting their VXML from the local LAN (maybe the same rack!). If the call is suddenly getting VXML from the UK and you need the fetchaudio stuff to make the user experience viable, you need to be able to not have this kick in on local VXML hits. You may be able to make this work - or you may not.


Cisco have a document on CVP over long latencies.


Having the call use CVP resources in the UK is the issue - not getting the call to the agent in the US. That part is fine.


When I said "India needs more than that", the above is exactly what I referred to.


Regards,

Geoff

ganeshraghuram Mon, 06/21/2010 - 15:50
User Badges:


You already told us once that this was not a mission critical call center - hence the absence of  that second subcriber. Don't overbuild it.



Yes not a mission critical with 20 and 10 agents across sites.However there is goin to be a presence of CCM Sub Secondary at the VA end also.



I hear what you are saying about "follow the sun". Dealing with CVP over long latencies requires a bit of fancy footwork with "fetch audio" on the gateway and you haev to do this anyway - because those trunks in India on the gateway there are fetching their VXML from London. So you need to get this right.



Sure we have got this right.Initially there was only US and HYD sites but due to high latency problem UK site has been introduced.The latency now is much below 200 msec (reaches 100 msec) even during the peak hours.The bandwidth quoted too based on the vxml pages exchanged between the server and Gw is sufficing.


http://www.cisco.com/en/US/docs/voice_ip_comm/cust_contact/contact_center/customer_voice_portal/srnd/7x/cvp_netw.html#wpxref30000



Having the call use CVP resources in the UK is the issue - not getting the call to the agent in the US. That part is fine.

When I said "India needs more than that", the above is exactly what I referred to.


At Hyd IND the customer just has a very small spacing with 10 agents seater call center.There is no data center available to.This was the reason for us to keep all the CVP servers in UK.The phones too presently are getting registered from UK UCM Cluster.


Geoff the UK Servers still does function for its primary role in servicing Hyd agents.Just an add-on to it would be too also use it from US in case of when they are not used by Hyd.

Bandwidth is not an issue. CVP can deal with being exposed over low latency connections as long as you make special provision on the gateway to use fetchaudio or the user experience suffers.

Sure we have got this right.Initially there was only US and HYD sites but due to high latency problem UK site has been introduced.The latency now is much below 200 msec (reaches 100 msec) even during the peak hours.


Then what are you asking? You don't appear to be listening to what I say.


Regards,

Geoff

ganeshraghuram Mon, 06/21/2010 - 18:45
User Badges:

My questions are simple


Taken from my first post...


1 In a case when my CVP 1(FL) is down and wan links are down at the same time which makes CVP 2(VA) of no use can i transfer calls from FL Gw to CVP in UK(Primary) and play some outage message and transfer the call to voice mail in UK.

This same scenario can also trigger in a case when my ports in FL CVP are full and i want the calls to be send to UK CVP.Can i use CVP Primary (UK) for servicing calls from US(FL) and IND(Hyd) if needed toghether.Should A2Q object this?


Which i can from the post.


2 Assuming i can use the above condition and transfer the calls to CVP (UK)(PG2)  from VG (FL) )and play the self service and then can i have the calls transfered to CCM (FL) (PG1) which sits on a diffrent PG box to have an agent transfer.


Technically it would be posible but would TAC disapprove it latter

Actions

This Discussion