Cisco Cisco Unified Contact Center Enterprise 9.0(1) Folheto

There is no impact to the agents, calls in progress, or calls in queue because the agents stay 
connected to their already established CTI OS Server process connection. The system can continue 
to function normally; however; the PGs will be in simplex mode until the private network link is 
restored.

If the two private network connections are combined into one link, the failures follow the same path; 
however, the system runs in simplex mode on both the Call Router and the Peripheral Gateway. If a 
second failure were to occur at that point, the system could lose some or all of the call routing and ACD 
functionality.

Prior to the 7.5(10) Maintenance Release, duplexed Peripheral Gateways (PGs) failed over to side A 
when the private network was lost between PG side A and side B. For example: 

PG side B is active (has active PIMs, PGAgent) 

PG side A and side B lose the private network 

PG side B unconditionally fails over to side A

This failover caused unnecessary disruption to the contact center, especially to Unified CCE, because 
ACD restarts and all agents need to sign back in, thus losing reporting and call control. This failure 
condition is more likely to occur when the PGs are split over the WAN.

This enhancement introduced in the 7.5(10) Maintenance Release uses a side-weight approach, which is 
similar—but not identical—to the device majority logic. Router component uses this majority logic.

In the case of PGs, the OPC process uses a weighted value based on the number of active components 
to decide the weight of both sides. This weight value determines which PG side takes over during a PG 
private link outage, so that the disruption to the contact center is minimized.

The enhanced OPC process dynamically calculates and maintains the weights for both the sides. OPC 
also keeps the MDS processes on both sides updated at all times with the cumulative device weights for 
both sides. This allows MDS on both sides to take appropriate action in case of a private link failure.

The PG predetermines the individual process of device weights on the basis of the impact to the contact 
center in case of failover. For example, Agent PIMs have a higher device weight than VRU PIMs and 
CTI Server, because if the Agent PIM fails over, the system takes longer to recover and the disruption 
to contact center is greater. The device weights for individual processes are not configurable.

The visible network in this design model is the network path between the data center locations where the 
main system components (Unified CM subscribers, Peripheral Gateways, Unified IP IVR/Unified CVP 
components, and so forth) are located. This network is used to carry all the voice traffic (RTP stream and 
call control signaling), Unified ICM CTI (call control signaling) traffic, as well as all typical data 
network traffic between the sites. In order to meet the requirements of Unified CM clustering over the 
WAN, this link must be highly available with very low latency and sufficient bandwidth. This link is 
critical to the Unified CCE design because it is part of the fault-tolerant design of the system, and it must 
be highly resilient as well:

The highly available (HA) WAN between the central sites must be fully redundant with no single 
point of failure. (For information regarding site-to-site redundancy options, refer to the WAN 
infrastructure and QoS design guides available at 

.) In case of 

partial failure of the highly available WAN, the redundant link must be capable of handling the full 
central-site load with all QoS parameters. For more information, see the section on 

.