User's Manual

26 Overview

Nortel Networks Confidential

Reliability and fault tolerance

The SIP Application Module provides reliability and fault tolerance

through multiple SIP Application Modules deployed in an N+M

active-standby configuration.

Note: The supported active/standby configurations include:

• a 1+1 configuration (one active plus one standby server), which

is the most basic reliable configuration

• an N+M configuration of up to four servers (the sum of N plus M

should not exceed 4)

— a 2+1 (2 active and one standby)

— 2+2 configuration

— 3+1 configuration

To accomplish this, all the servers in a reliability group are configured

with the same set of NSDs. This gives the standby server the

information it needs in case an active server fails. Each server in the

group transmits messages indicating its current state. Other servers

respond with their current states, including the NSD activated on them.

An initializing server configures itself with one of any inactive NSDs. If

all NSDs are active, the initializing server becomes the standby. This

prevents conflicts where more than one server is activating

simultaneously.

Before activating, the server determines whether it is isolated from

critical network resources defined through provisioning. If any of the

resources cannot be reached, the server cannot activate and raises an

alarm. The alarm clears when the resources become available.

When there are two or more active servers, the group is called a cluster.

You can configure both the N+M strategy and the cluster at the

Transport Management tab in step 22 in the Configuration chapter.

When one of the active SIP Application Modules fails, the passive

Module takes over the IP address. The passive Module has now

become active and assumes the responsibilities of the failed Module.

When this occurs, any sessions already in the active state remain up.

This means that calls that have already been established continue and

the parties maintain voice path. Any future requests during that session,

however, fail (for example, Hold, Retrieve, and Web Pushes) since the

session information is no longer available. Any sessions that were not

in the active state before the failover are lost. The originating clients of