Opinions about CAC in multiple clusters with multiple locations each.

Answered Question
Mar 28th, 2008
User Badges:

Hi, my company operates in several countries and we are designing IP Telephony and VoIP to leverage our global WAN usage and slowing adopting IPT. Right now we are on two locations in Sao Paulo, one of them on what we call Full IPT site (7911 and 41 phones, some IP Comm and a 2851 with two E1 to PSTN) and the older one using traditional PBX linked to two 2851 for redundant connection to PSTN and Alcatel PBX. We intend to expand this model across to 8 more sites in Brazil, two using full IP telephony and the others having the voice gateway intermediating call between PBX and PSTN, also handling on net calls in the global WAN for intracluster calls. Voice gateways are MGCP controlled (because of FAX issues on using ISDN) with H.323 fallback and SRST.

The project also includes a new cluster in Buenos Aires controlling this site and some other sites in southern South America and another cluster in Santiago for 6 cities in Chile and some others in andian countries.

I am not quite sure on how to deploy CAC between sites in different clusters. Locations cannot be used between clusters, if I'm not wrong. Also gatekeeper would be able handle CAC between H.225/GK controlled clusters (we intend to have all 3 clusters trunking with 2 GKs). But since the how global network is MPLS and each site has direct connection to the others, the gatekeeper wouldn't be able to know to each site in a cluster the bandwidth is allocated.

It looks like RSVP can be used to control admission between all sites because they can be used in intercluster trunks. Since we cannot control MPLS edge gateways as they are managed by the SP (other than telling them to provide a specific LLQ bandwidth), we would be using the voice gateways with two gigabit ethernet interfaces between SP routers and LAN swiches (aggregating a few more network addresses in this task) since SP routers can't do RSVP.

Have someone tried a setup on a network like that? Would that be a bad design (considering we could not deploy just one cluster across whole South America since some sites have more than 720 people and to have more redundancy) or any big caveats? I would assume that by using RSVP in CUCM6.x for that there would be no need of using locations for CAC or could it be using integrated with RSVP?

I'd be glad to hear comments and ideas. Regards.

Correct Answer by matthewpage about 9 years 3 months ago

Hi


For the ICT trunks between the clusters you would have to have IP-IP GWs if you wanted to use RSVP. It will not work with just GKPs but of course if your using just gkps couldnt you put each cluster into a zone then restrict the bandwidth to each one?? I guess you would also need a star topology for that though.


Cant you enable the rsvp on the voice gateway before it hits the SP edge? If RSVP is not enabled on a device it becomes transparent to the RSVP so it will work even though the SP doesnt support it.


If you going to use RSVP best to use it in intser/diffserv mode as this allows you to apply your own Qos and queuing appose to RSVP doing it all.


I have done a similar setup with 4 clusters using Gatekeepers with IPIPGws which are running RSVP. One cluster had 19 remote sites running FR and MPLS all RSVP enabled. I didnt have any problems other than playing around with how much bandwith to allocated to each interface.


If you need any RSVP configs or any help with that let me know.


hope some of that helps


Matt


  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 5 (4 ratings)
Loading.
Correct Answer
matthewpage Fri, 03/28/2008 - 13:50
User Badges:
  • Bronze, 100 points or more

Hi


For the ICT trunks between the clusters you would have to have IP-IP GWs if you wanted to use RSVP. It will not work with just GKPs but of course if your using just gkps couldnt you put each cluster into a zone then restrict the bandwidth to each one?? I guess you would also need a star topology for that though.


Cant you enable the rsvp on the voice gateway before it hits the SP edge? If RSVP is not enabled on a device it becomes transparent to the RSVP so it will work even though the SP doesnt support it.


If you going to use RSVP best to use it in intser/diffserv mode as this allows you to apply your own Qos and queuing appose to RSVP doing it all.


I have done a similar setup with 4 clusters using Gatekeepers with IPIPGws which are running RSVP. One cluster had 19 remote sites running FR and MPLS all RSVP enabled. I didnt have any problems other than playing around with how much bandwith to allocated to each interface.


If you need any RSVP configs or any help with that let me know.


hope some of that helps


Matt


rodmont74 Tue, 04/01/2008 - 16:11
User Badges:

Thanks a lot Matt. I guess RSVP with intserv/diffserv fits very nice with this deployed topology. Yes, the star topology would allow us to use GKs but the network is P2P. Add to that, most sites are going to continue TDM PBXes using VoIP to on net calls. Hopefully, the full IPT sites will work very well and added IPT feature compel leaders to start other sites migration.Well, good challenge anyway deploy it. Hope learning experience helps me achieving my CCVP!

Regards.

matthewpage Wed, 04/02/2008 - 01:27
User Badges:
  • Bronze, 100 points or more

Yeah RSVP will be fine between the central and already roled out IPT sites.


I was thinking for the sites that are still running PBXs you must have a VOIP gateway already inplace to handle offnet calls over the wan. You could upgrade this to a IPIPGW/GKP/Gateway all on one box. Then have a central GKP at your central site. This will enable to you run RSVP between all the sites.


HTH


Matt



rodmont74 Fri, 04/04/2008 - 06:32
User Badges:

Thanks for the ideas. Looks like we have some variations to work out. I have prepared a small diagram of how it could work (forgive my awful drawing skills).



matthewpage Sun, 04/06/2008 - 07:26
User Badges:
  • Bronze, 100 points or more

Hi Rodrigo


Im dont quite understand the call flow you have plotted out.


If you were making a call from Rio de Janeiro to Assuncion wouldnt call be direct between the two sites? Or are you showing the call setup as well. What type of gateways are you using at the other sites?


Nice drawing by the way!


Matt

rodmont74 Mon, 04/07/2008 - 04:07
User Badges:

Thanks for point it up Matt. Yes, it's setup flow as I supose it will work. Call flow will be direct between the sites, something I need to add later to the draft.

Gateways are going to be 3825 mostly and some 3845 for large offices to support big SRST deployments. All sites in Brazil, Chile and Argentina will go with ISDN over E1 and some andean sites on FXO ports to the PSTN or current old PBX (depends on PBX capabilities and some issues that may arise if voice gateways are placed between PSTN and PBX using analogic connections.


Regards.

matthewpage Mon, 04/07/2008 - 06:37
User Badges:
  • Bronze, 100 points or more

OK looking at the diagram again i understand what you were saying. So the RSVP would flow from the (cluster1)SRST ->IPIPGW - IPIPGW->SRST(Cluster2).


The call manager isnt aware of the RSVP between clusters as it doesnt support it. That is when the IPIPGWs come into play.


If you have sites that are using H323 gateways i think you can enable RSVP on the dial-peer and dont need a IPIPGW in the way.



This is going to be a big deployment and i would really recommend getting this set up in a lab first.


Let me know if you need help with any of the configurations.

rodmont74 Mon, 04/07/2008 - 09:09
User Badges:

Thank you very much, excelent information Matt. More detail on draft I need to work out as well.


I understand the new IPIPGW role could be in one of the GK routers and possibly secondary on secondary GK.


I'm about to do GWGK exam on 22nd, but I have not experience on GK and IPIPGW. Unfortunately this project won't go as fast as I'd like. On the good side I should gain experience on those two topics when service provider is deploying this setup.


I do agree this is huge, much complicate than going on on a single cluster deployment but hopefully it should be scalable globally.


I'll post more info here as I get news from the development.


Best regards.

matthewpage Mon, 04/07/2008 - 10:44
User Badges:
  • Bronze, 100 points or more

Your understanding is right with the IPIPGW role. You can have one box that acts as a GKP and IPIPGW. This will save you a bit on hardware costs.


RSVP is very scalable so you shouldnt have any problems there.


Unfortunatly you dont cover RSVP in GWGKP i dont actually remember covering it on any of the CCVP exams.


matt


Actions

This Discussion