Cisco Cisco IP Contact Center Release 4.6.1 Design Guide

The terms public network and visible network are used interchangeably throughout this document.

In clustering over the WAN with Unified CCE, there should be an isolated private network connection 
between the geographically distributed Central Controller (Call Router/Logger) and the split Peripheral 
Gateway pair to maintain state and synchronization between the sides of the system.

To understand this scenario fully, a brief review of the ICM Fault Tolerant architecture is warranted. On 
each call router, there is a process known as the Message Delivery Service (MDS), which delivers 
messages to and from local processes such as router.exe and which handles synchronization of messages 
to both call routers. For example, if a route request comes from the carrier or any routing client to side A, 
MDS ensures that both call routers receive the request. MDS also handles the duplicate output messages.

The MDS process ensures that duplex ICM sides are functioning in a synchronized execution, fault 
tolerance method. Both routers are executing everything in lockstep, based on input the router receives 
from MDS. Because of this synchronized execution method, the MDS processes must always be in 
communication with each other over the private network. They use TCP keep-alive messages generated 
every 100 ms to ensure the health of the redundant mate or the other side. Missing five consecutive TCP 
keep-alive messages indicates to Unified ICM that the link or the remote partner system might have 
failed.

When running duplexed ICM sides as recommended for all production system, one MDS will be the 
enabled synchronizer and will be in a paired-enabled state. Its partner will be the disabled synchronizer 
and is said to be paired-disabled. Whenever the sides are running synchronized, the side A MDS will be 
the enabled synchronizer in paired-enabled state. Its partner, side B, will be the disabled synchronizer 
and paired-disabled state. The enabled synchronizer sets the ordering of input messages to the router and 
also maintains the master clock for the ICM system.

If the private network fails between the Unified ICM Central Controllers, the following conditions apply:

The Call Routers detects the failure by missing five consecutive TCP keep-alive messages. The 
currently enabled side (side A in most cases) transitions to an isolated-enabled state and continues 
to function as long as it can communicate to at least half of the configured number of Peripheral 
Gateways (PGs). The disabled side (side B in this example) checks that it can communicate with a 
majority of the PGs, or half of the configured PGs plus one (half + 1). If the disabled side (B) cannot 
communicate with a majority (half + 1) of the PGs, then it transitions to an isolated-disabled state 
and side A runs in simplex mode. (Noted that any PG with a connected DMP path, either active or 
idle, is counted in this majority calculation for both side A and side B.) If the B side does have device 
majority (can communicate to half + 1 of the configured PGs), then it transitions to a testing state 
and initiates the Test Other Side (TOS) procedure to determine if it should become enabled.

The disabled synchronizer (side B in this case) sends TOS messages through each PG in sequential 
order to Router A. One successful message from any PG, communicating that Router A is still 
enabled, will cause Router B to transition to the isolated-disabled state and make itself idle. Its 
mated Logger B will also idle itself. All the Peripheral Gateways realign their active data feed to the 
active Call Router over the visible network, with no failover or loss of service because the PGs do 
not disconnect when they realign.

If all PGs reply that side A is down during the TOS procedure, then side B promotes itself to 
isolated-enabled state and runs in simplex mode.