Cisco Cisco IPCC Web Option Design Guide

The Agent PG uses the Unified CM CTI Manager process to communicate with the Unified CM cluster, 
while a single Peripheral Interface Manager (PIM) controls agent phones anywhere in the cluster. The 
Peripheral Gateway PIM process registers with CTI Manager on one of the Unified CM servers in the 
cluster, and the CTI Manager accepts all JTAPI requests from the PG for the cluster. If the phone, route 
point, or other device being controlled by the PG is not registered to that specific Unified CM server in 
the cluster, the CTI Manager forwards that request via Unified CM SDL links to the other Unified CM 
servers in the cluster. There is no need for a PG to connect to multiple Unified CM servers in a cluster.

Although the Agent PG in this document is described as typically having only one PIM process that 
connects to the Unified CM cluster, the Agent PG can manage multiple PIM interfaces to the same 
Unified CM cluster, which can be used to create additional peripherals within Unified CCE to separate 
registration for CTI route points and phones into two different streams. Using two PIMs for one cluster 
is required only when a customer has a large number of CTI Route Points in the Unified CCE 
Configuration. A single PIM can register approximately five CTI Route Points per second. To reduce 
initialization and failover times, you should use a second PIM when there are more than 250 Route 
Points. It is recommended that you add another PIM to the Unified CCE PG to handle only the CTI Route 
Points. With this model, all agent control is contained within a distinct PIM, and all CTI Route Points 
are contained on a second PIM. This enables the CTI Route Point PIM to register the Route Points and 
bring the system online before all of the agent phones are registered. This model applies only to Unified 
CCE models. Note that system Unified CCE allows only one Unified CM PIM.

The typical process for the startup of a PIM is to register with the peripheral for all of the monitored 
objects, such as CTI route points, dialed numbers, agent phones, or device targets. The PIM will not go 
active until all of these objects are registered and monitored properly in the PIM connection. 

Many Unified CCE implementations have only a small number of CTI route points, while having 
numerous device targets to register. By separating these two configuration objects across two PIMs, the 
CTI route points can be registered and used for routing long before all of the agent phones or device 
targets are registered and active. The concept is to have the CTI route points available to Unified CCE 
immediately upon PG/PIM startup and failover, thus allowing Unified CM to process these route 
requests and start to handle calls even if the agent phones are not all registered. Callers can be put into 
queue quickly or played a message rather than having the default forward-on-failure processing in 
Unified CM take over the calls if the PIM is not ready to accept the incoming calls. (However, this 
functionality is not available in System Unified CCE, which allows only one PIM per PG. Only 
traditional Unified CCE configurations allow for multiple PIM configuration.)

Duplex Agent PG implementations are highly recommended because the PG has only one connection to 
the Unified CM cluster using a single CTI Manager. If that CTI Manager were to fail, the PG would no 
long be able to communicate with the Unified CM cluster. Adding a redundant or duplex PG allows the 
Unified ICM to have a second pathway or connection to the Unified CM cluster using a second CTI 
Manager process on a different Unified CM server in the cluster.

The minimum requirement for Unified ICM high-availability support for CTI Manager and Unified 
IP IVR is a duplex (redundant) Agent PG environment with one Unified CM cluster containing at least 
two subscribers. Therefore, the minimum configuration for a Unified CM cluster in this case is one 
publisher and two subscribers. This minimum configuration ensures that, if the primary subscriber fails, 
the devices will re-home to the secondary subscriber and not to the publisher for the cluster. (See 

.) In smaller systems and labs, Cisco permits a single publisher and single subscriber, which 

means if the subscriber fails, then all the devices will be active on the publisher. For specific details about 
the number of recommended Unified CM servers, see 

.