Cisco Cisco IPCC Web Option Leaflet

In a network where Hot Standby Router Protocol (HSRP) is deployed on the default gateways that are 
configured on the Unified CCE servers, follow these recommendations:

Configure the HSRP hold time (plus the associated processing delay) to be lower than five times the 
heartbeat interval (100 ms on the private network and 400 ms on the public network) in order to 
avoid ICM private network communication outage during HSRP active router switch-over.

With convergence delays that exceed private or public network outage notification, HSRP 
failover times can exceed the threshold by which network outage detection is made, thus 
causing the enterprise system to complete a failure and recovery phase. If primary and 
secondary designations are made in the HSRP configuration and the primary path router 
fails to the secondary side, HSRP will subsequently reinstate the primary path when 
possible, thereby leading to a second private network outage detection.

For this reason, configured HSRP convergence delays that approach 500 ms for the private network 
and 2 seconds for the public network are best not configured with primary and secondary 
designations to avoid the start-path reinstatement mentioned above. On the other hand, convergence 
delays that can be configured below the detected threshold (which thus render an HSRP failover to 
be transparent to the application) do not mandate a preferred path configuration. This approach is 
preferable. Cisco recommends keeping enabled routers symmetrical if path values and costs are 
identical. However, if available bandwidth and cost favor one path (and the path transition is 
transparent), then designation of a primary path and router is advised.

The ICM fault-tolerance design requires the private network to be physically separate from the 
public network, therefore HSRP should never be configured to fail-over the private network traffic 
to the public network link, or vise versa.

The bandwidth requirement for ICM should be guaranteed anytime with HSRP, otherwise the 
system behavior is unpredictable. For example, if HSRP is initially configured to do load sharing, 
there should still be sufficient bandwidth for ICM on the surviving links in the worst-case failure 
situations.

Cisco Unified Communications Manager (Unified CM) 5.0 introduces support for Resource 
Reservation Protocol (RSVP) between endpoints within a cluster. A protocol for Call Admission Control 
(CAC), RSVP is used by the routers in the network to reserve bandwidth for calls.

To calculate bandwidth usage before RSVP was introduced, it was necessary for each Unified CM 
cluster to maintain local counts of how many active calls traversed between locations. If more than one 
Unified CM cluster shared the same link, it was necessary to dedicate bandwidth for each cluster, leading 
to inefficient use of available bandwidth.

RSVP solves this problem by tracing the path between two RSVP agents that reside on the same LAN 
as the phones. The RSVP agent is a software media termination point (MTP) that runs on Cisco IOS 
routers. The RSVP agents are controlled by Unified CM and are inserted into the media stream between 
the two phones when a call is made. The RSVP agent of the originating phone will traverse the network 
to the RSVP agent of the destination phone, and reserve bandwidth. Since the network routers keep track 
of bandwidth usage instead of Unified CM, multiple phone calls can traverse the same RSVP controlled 
link even if the calls are controlled by multiple Unified CMs.