JTAPI CTI Connection capacities for multiple providers

[perlocper]

Hi all,

As I read from the documents, each cluster can support a maximum of 40,000 CTI connections with Unified CM 8.6(1) and later releases.

In my design, I want to create providers for each of the end users. So there will not be any app users in the system. Each end users will monitor themselves and get notifications only for their actions. I wonder how this will affect the capacity of the system.

Let's say that, i have 20000 devices-hard phones (with number of lines is 1) in the system which means I need to create 20000 providers for each of them. Since the total of them is 40000, I assume it should work.

I also read that a phone with 5 lines counts as 5 CTI devices. So if the number of lines is 2 for each devices, then does it mean that I can only monitor half of the actual hard phones? (10000 devices) So do I need to create only 10000 providers in that case, or should it be 20000 (10000 device * number of lines)?

Thanks.

[mpotluri]

Collaboration sizing tool would help in estimating the capacity of your cluster. If you haven't tried i recommend using: http://tools.cisco.com/cucst

Since each user is a CTI connection and a server pair supports 10,000, you will need 4 pairs of subscribes (8 subscribers) to handle 40,000 users.  A pair is recommend to handle any fail over scenarios. In your example of 20,000 users you will need 4 subscribes.

Starting from 8.6.2 CTI supports 5 CTI-enabled lines (unique DN or shared DNs) per device without requiring additional resources. So 2 lines on a device is still 1 CTI resource. HTH

[perlocper]

Thanks for the explanation Mohan. But I am still confused about the numbers. Can you please look at the following table and check the numbers if they are correct? I considered different scenarios in order to understand Cisco capacity limits better.

	Number of devices that will be monitored	Number of providers that will be used for monitoring	Number of total CTI connections	Number of subscribers
Scenario 1:	10,000	10,000 (I create a provider for each of the users)	20,000	2 pair (4 subscribers)
Scenario 2:	20,000	20,000	40,000	4 pairs (8 subscribers)
Scenario 3:	100,000	100,000	200,000	20 pairs (is it supported? Is 8 the max number of pairs? How can the model be in order to support 200,000 CTI connections?)

As far as I understand, these numbers will not change according to the number of lines on devices (up to 5 DNs on a device).

Thanks.

[mpotluri · Accepted solution]

'Number of providers that will be used for monitoring' should be same as CTI connections. One provider will result in 1 CTI connection. Each user will have 1 provider which will result in 1 active CTI connection.  So those 2 columns should have the same number.

Since 1 node can support 10k providers with 1 device each (10k CTI connections) and a second one is a failover node we need 2 nodes (1 pair) for 10k users.

20k users will need 2 pair (4 subscriber nodes)

Normal cluster can have 8 subscriber nodes, so 40k users is the max we can support on a typical cluster.

Mega clusters (Cisco Collaboration System 10.x Solution Reference Network Designs (SRND) - Call Processing [Cisco Unified Communication…) can have 80K devices. So if all phones are end user phones you can go up to 80k users in a mega cluster. HTH.