I'm a newbie to CAC so would greatly appreciated if someone can shed some light on this.
What's the main difference between the use of locations VS use of GK to limit bandwidth to sites? When will you use which method and is there any pros and cons on each method?
GK is needed only when you have a distributed deploykment model where you have multiple callmanager clusters or callmanager at central site with cme at remote sites. GK will help control how many calls are going from one cluster to another cluster.
Locations are used in a centralized call processing model where remote sites are running off of the centralized call manager. You may define regions to define what codec will be used across the wan. But still there needs to be a way to limit the number of g729 calls that go across the wan. In this case there is no need for GK because all phones are part of the same cluster so the Callmanager can do dial plan resolution by itself. You define locations in this case for each remote site. The main site can be considered as having unlimited bandwidth. You can define one location for each remote site and specify the number of calls that can be placed from that site to any other site. Calculations are done for g711 calls at 80kbps per call and 24 kbps for g729. So if you are using g729 across the wan and you want to limit to 3 calls from a remote site, you can set the location bandwidth for that site to 24*3 = 72kbps.
Hope that makes sense!
PS: please remember to rate posts!
Thanks for your help, Sankar! There is still something i don't understand and hopefully you can shed some light on this.
i) On the CCM location page, since we can only specify the "total bandwidth" coming in and out of the whole location (but not individually from location-to-location), does it mean when we have a 3 tier topology, the calculations for the bandwidth may be incorrect?
e.g. In a 3 tier model
HO (CCM cluster) --- Hub Office (site A), where site A (hub) is connected to 2 remote offices (site B & site C)
Now say the locations are like these:
Site A (hub) - location of 10 calls
Site B (RO) - location of 10 calls
Site C (RO) - locaiton of 15 calls
And assume Site A places 5 calls to site B while site C also places 5 calls to site B simultenously. So site B has used up all the bandwidth in the location and can't place any calls to HO.
My understanding is because Locations are not supposed to support these kind of 3 tier scenario, there is no workaround to allow site B to call HO while it is receiving 5 calls from site A and 5 calls from site C. Am i correct in saying locations doesn't work for this scenario?
BUT, there is always a but =) if this is a GK controlled trunk scenario, where it becomes 4 seperate CCM clusters, can we fix this?
ii) When locations are used, when that "particular location" ran out of bandwidth, i noticed the phones will say "Not enough bandwidth". Does this msg also shows if the GK is used instead of bandwidth?
CAC location based and GK based are solutions for hub and spoke topology only. With the 3 tier topology you will never get completely functional, you can do your best estimating the bandwidth for the tier 2 site. Look for RSVP to resolve your isses available in CCM 4.2.
One other design difference to consider...
Both CCM and GK can only account for calls they are aware of.
CCM CAC will account the bandwidth fo all calls it knows about from location to location. This can include IP Phone to IP Phone, Phone to Gateway, Gateway to Gateway, Intra-cluster, and Inter-cluster calls, as long as CCM is in the control path.
Gatekeepers, on the other hand, can only control bandwidth for calls it is aware of. This will include H323 gateways and Terminals that are registered with the GK and use it for a particular call. Examples include cluster to cluster, or cluster to gateway. It will not include Phone to Phone or intra-cluster.
Please let me know if this helps by rating the post.
For GK, what are Intra-cluster calls? Do you mean not all sites will have it's own cluster, and hence certain sites are hub and spoke and the calls between that hub and spoke is intra-cluster? If so, can we use location along side alongside with GK CAC?
For phone to phone, what do you mean by that? If each site has their own cluster, do we still have to worry about phone-to-phone?
Sorry I wasn't clear. I'm trying to think of clearer verbage and an example to illustrate. I need to post this in two parts since it is a bit long.
H323 endpoint = When configured to use a Trunk or H323 Gateway, the CCM becomes an H323 endpoint. H323 gateways are also H323 endpoints.
What I mean by Intra-cluster calls is any call where both endpoints are controlled by a single CCM cluster. Think of a centralized deployment with a single CCM cluster.
Also, since the GK is only asked for RAS between H.323 endpoints, it cannot provide direct control over MGCP gateways, IP Phones, nor SIP endpoints when those devices are operating from the same CCM cluster. In such cases, CCM Location settings must be used to provide CAC.
In situations where there is more that one CCM cluster, and/or they need to share multiple h323 gateways, and/or the gateways are at remote locations where CAC is important, then the GK is of great use.
The GK would be the device(s) where you consolidate your dialplan. It could also keep track of bandwidth for calls between different CCM clusters and h323 gateways, as well as track bandwidth used between these h323 endpoints.
The GK sees each CCM cluster's trunk, and each gateway, as a separate H.323 entity. It has no visibility into callmanager locations. It doesn't know where phones are located, and it doesn't know about the existence of MGCP gateways.
Should you use both GK and CCM CAC? It depends. The issue is that there is no mechanism to sync the bandwidth utilization tracked on the GK with that which is tracked by CCM Locations.
Imagine an enterprise with two subsidiaries. Each subsidiary has its own CCM cluster at its respective HQ such that there is HQ-A and HQ-B. And imagine that enterprise has a remote WAN connected site connected to HQ-A via 128Kbps WAN, with several IP Phones, and a single h323 gateway (SiteC). To save money, you have been told to configure toll bypass such that calls destined for the same local area code (512) as SiteC use the SiteC gateway regardless if the call originates in HQ-A, HQ-B, or any other enterprise location. CCM Location based CAC is configured for the SiteC gateway in both clusters and is set to 128k to reflect the link speed. Suppose for argument's sake that a G729 call actually consumes 24kbps on this particular WAN link.
Given this enterprise, suppose these two call flows:
1) HQ-A IPPHone users place three G729 calls to the 512 area code. These calls are routed to the SiteC gateway for toll-bypass.
2) Before anyone at HQ-A has hung up, HQ-B IPPHone users place three additional G729 calls to the 512 area code. These calls use the same gateway at SiteC, also for toll-bypass.
How much bandwidth is allocated in each CCM cluster for that single 128kbps WAN circuit? How is the quality of those calls?
If HQ-A and HQ-B have the SiteC gateway configured as an h323 gateway, then each cluster will allocate 72Kbps for its own calls. However 6 calls are in progress, so the voice path requirement is 6 x 24 = 144kbps. Packets are getting dropped, and depending on QoS, routing protocols may break, etc. In any event, call quality is very bad.
In fact, each CCM (HQ-A or B) could potentially try to allow 5 calls to the SiteC gateway (5 x 24kbps = 120) leading to an extremely oversubscribed link (10 x 24kbps = 240kbps).
If we put a GK in the mix, by making both CCMs hand controll of the SiteC bound calls to a GK controlled trunk rather than the SiteC gateway, the GK can monitor the total bandwidth used by the Gateway and limit the number of calls to 5 regardless of the CCM cluster making the call.
The caveat here is the IP Phones at SiteC. They are controlled by HQ-A's CCM. And they may make calls to other enterprise sites over this same WAN link. So at any time, you may not have 128k available for toll-bypass calls. CCM Locations are the only way to control the max number of IP Phone calls out of SiteC. So you could have a situation where both Location and GK CAC is used.
The art is in balancing the max number of SiteC IPPhone calls with the max number of toll-bypass calls because the GK and CCM don't share utilization information.
Let me know if this helps by rating the post.
I have a few Qs regarding your scenario.
How do we register the Gateway at siteC straight to the GK? Do we define it a zone like how a CCM would register with the GK?
Back to the CCM CAC location scenario, so on CCMA at HQ-A, do i put all SiteC's phone and Gateway into a new location called site C with 128kbps bandwidth, and then put all HQ-A phones into is own location called site HQ-A with unlimited bandwidth?
And do I have to use BAT to assign locations of each individual phones on both Site C and HQ-A coz there is no trunk that we can set location to?
If Site C is only connected to WAN to HQ-A and controlled by HQ-A's cluster, how can we limited that to 128k in location CAC at HQ-B cluster? If we define site B as a location on HQ-A CCM cluster to 5 calls, doesn't it then restrict the calls of HQ-B to HQ-A and Site C a total of 5 calls only?
I have seen before when the location is out of bandwidth, the user can see on their IP phones displaying "Not enough bandwidth". Does a similar mechanism exist for GK solution?
2) I would place it in a separate local zone so I can specify bandwidth for that zone. Problem is, that IPPhones at SiteC will be constrained by that same bandwidth limitation - even though this is not desired.
3) You would put SiteC Gateway & Phones into a single SiteC location. Assign 128Kbps or less bandwidth to that Location. SiteC devices will have unlimited bandwidth to talk to one another. Other locations will be constrained to 128k.
4) BAT would probably be easiest approach. But even if you did assign a location to the trunk, you won't want to constrain IP Phones that are physically at SiteC to 128k since their RTP traffic doesn't need to go across the WAN to get to the gateway.
5) I'm not sure I understand the question, but I'll take a stab. My understanding is that CCM checks bandwidth against the destination's Location bandwidth pool, not the source's. So you could use an inter-cluster trunk (ICT) between HQA and HQB with HQ-A controlling the call. That would be a valid workaround to not using the GK. I didn't include it earlier since we were discussing how things work and I don't know much about your environment.
6) I've never tried to use AAR for such a purpose. However, since the GK is tracking that bandwidth anyway, and there might be alternate gateways that the GK could choose to place the call, I typically will leave the location set to unlimited for the GK controlled trunk. I would handle failover by use of Route-groups/Route-lists to alternate gateways.
All in all, most folks are inclined to use CCM Locations in a centralized deployment unless some specific requirement forces them to de-centralize.
See the following for pros/cons of the various deployment approaches:
Don't forget to rate the posts...