Collaboration Infrastructure and Design
Domain 1: Collaboration Infrastructure and Design
This domain tests how you size, place, and harden the components that carry registration, database replication, media, and dial-plan policy. Expect design trade-offs around CUCM clustering, redundancy, QoS, virtualization, and address plan, not isolated menu recall.
Core rules
- CUCM uses a publisher-subscriber model. The publisher is the database source of truth; subscribers hold replicated copies and continue call processing during a publisher outage, but administrative changes require the publisher.
- NTP consistency is mandatory for database replication and certificate validation; DNS is optional for basic CUCM operation but becomes important for SIP, Expressway, and service discovery.
- Device pools bind CUCM groups, regions, locations, date/time groups, and SRST references. A CUCM group gives registration redundancy, not call survivability; SRST or a local gateway provides fallback.
- Regions define the codec between sites; locations define the bandwidth budget for call admission control. Do not use one to solve the other problem.
- Mark voice media EF, video AF41, and call signaling CS3 at the trust boundary; include L2/L3 overhead in CAC calculations.
- A globalized dial plan uses +E.164 as the canonical internal number, with translation patterns for local breakout and AAR for CAC failure.
Worked example
A branch has 200 phones and a 50 Mbps WAN to a centralized CUCM cluster. Intra-branch calls should use G.711; branch-to-HQ calls should use G.729. Create a branch region with G.729 to other regions and G.711 within the branch; create a branch location with 50 Mbps and subtract signaling and overhead; assign the branch device pool to that region and location; add an SRST reference for WAN loss; configure AAR to PSTN. Reasoning: the region answers which codec; the location answers how much bandwidth; SRST answers what happens when the WAN fails.
Traps
- Assuming a CUCM group provides SRST. It only provides registration failover.
- Forgetting NTP before certificate operations, causing replication and TLS failures.
- Mixing region and location roles, so CAC permits more calls than the WAN supports.
- Omitting transcoders or MTPs when a codec mismatch or DTMF interworking requires them.
- Designing DNS without correct SRV/A records and then blaming SIP registration.