5.1 Call Admission Control (CAC), Locations & AAR

5.1 Call Admission Control (CAC), Locations & AAR

In enterprise unified communications networks spanning multiple geographic sites, wide area network (WAN) links represent constrained and costly resources. While Quality of Service (QoS) mechanisms prioritize real-time traffic over data bursts, QoS cannot synthesize additional physical bandwidth. If unmanaged voice and video streams exceed the provisioned Priority Queue (PQ) capacity, every active call traversing the link suffers packet loss, jitter, and buffer overrun. Call Admission Control (CAC) solves this fundamental challenge by enforcing call admission limits at the signaling layer before media streams are permitted onto the network.

---

1. Call Admission Control (CAC) Rationale & Architecture

Quality of Service and Call Admission Control serve distinct, complementary purposes:

• Quality of Service (QoS): Implemented on routers and switches at Layers 2 and 3. QoS classifies, marks, and queues packets (e.g., placing voice RTP into Low Latency Queuing with DSCP EF). However, if 15 calls traverse a link sized for 10 calls, the priority queue overflows, causing tail drops across all 15 calls. QoS protects voice from data; it cannot protect voice from voice.
• Call Admission Control (CAC): Implemented at Layer 7 within call processing agents like Cisco Unified Communications Manager (CUCM). CAC tracks available bandwidth on defined network paths. When a link reaches its provisioned threshold, CAC rejects or redirects subsequent call setup attempts, guaranteeing that already-admitted calls maintain pristine voice and video quality.

+--------------------------------------------------------------------------------+
|                                 CAC DECISION FLOW                              |
|                                                                                |
|  Calling Party                    CUCM CallManager                Called Party |
|       |                                  |                             |       |
|       |--- 1. Off-hook & Dial Digits --->|                             |       |
|       |                                  |-- 2. Check Locations Pool   |       |
|       |                                  |      Is Bandwidth Avail?    |       |
|       |                                  |      YES: Deduct Bandwidth  |       |
|       |                                  |      NO:  Trigger AAR / Drop|       |
|       |                                  |                             |       |
|       |<-- 3a. Ringback Tone (Admitted) -+-- 3a. SIP INVITE (Admitted) ------->|
|       |                                  |                             |       |
|       |                                  |-- 3b. Bandwidth Exhausted --|       |
|       |<-- 3b. Fast Busy (No AAR) -------+                             |       |
|       |               OR                 |                             |       |
|       |<-- 3c. Reroute via PSTN (AAR) ---+-- 3c. Outbound PSTN Trunk ->|       |
+--------------------------------------------------------------------------------+

---

2. Location-Based CAC in CUCM

CUCM provides built-in Location-based CAC to manage WAN bandwidth consumption without requiring external RSVP signaling agents or gatekeepers.

The Hub_None Location and Topologies

• Hub_None: The default factory location in CUCM. Every newly configured device, gateway, and trunk belongs to Hub_None unless explicitly assigned elsewhere. By default, Hub_None has unlimited bandwidth allocated to all other locations.
• Hub-and-Spoke Topology: A central hub location (typically Headquarters or a primary Data Center) interconnects multiple remote branch spoke locations. Remote locations link directly to the central hub.
• Inter-Location Links: Administrators define logical links connecting pairs of locations. Each link specifies independent bandwidth limits for:
  1. Audio Bandwidth Pool: Measured in kilobits per second (kbps).
  2. Video Bandwidth Pool: Measured in kbps for standard desktop/room video.
  3. Immersive Video Bandwidth Pool: Dedicated pool for multi-screen telepresence systems, preventing high-bandwidth room systems from starving standard video or voice.

Bandwidth Deduction Calculations

CUCM calculates bandwidth reservations based on Layer 3 packet requirements (codec payload plus 40 bytes of IP/UDP/RTP headers at 20 ms packetization):

• G.711 ($\mu$-law / A-law): 64 kbps payload + 16 kbps L3 overhead = 80 kbps deducted per call.
• G.729 / G.729a: 8 kbps payload + 16 kbps L3 overhead = 24 kbps deducted per call.
• G.722 (Wideband): 64 kbps payload + 16 kbps L3 overhead = 80 kbps deducted per call.

The maximum number of simultaneous audio calls admitted across a link is determined by:

$N_{\text{calls}} = \left\lfloor \frac{\text{Allocated Bandwidth Pool (kbps)}}{\text{Codec Bandwidth Deduction (kbps)}} \right\rfloor$

For example, if an inter-location link has an audio pool of 240 kbps:

• With G.711: $\lfloor 240 / 80 \rfloor = 3$ concurrent calls.
• With G.729: $\lfloor 240 / 24 \rfloor = 10$ concurrent calls.

Video Call Bandwidth Allocation

A video call requires both an audio stream and a video stream. When a video call is initiated, CUCM deducts the audio portion from the Audio Pool and the negotiated video portion from the Video Pool. If the video pool has insufficient bandwidth but the audio pool has capacity, CUCM evaluates the device configuration to determine if the call may fall back to audio-only mode.

---

3. Enhanced Location CAC (ELCAC) in Multi-Cluster Networks

Since CUCM 9.0, Locations CAC is implemented as Enhanced Locations CAC (ELCAC): locations are joined by links that carry the bandwidth values, which lets CUCM model multi-hop topologies inside one cluster and, with Location Bandwidth Manager (LBM) replication, across several clusters.

ELCAC Architectural Components

1. Intercluster Communication: Each cluster's Location Bandwidth Manager (LBM) service replicates its location topology and bandwidth usage to the other clusters through LBM intercluster (hub) replication. The SIP intercluster trunks carry only the calls.
2. Shadow Location: When a call traverses an inter-cluster SIP trunk, standard Location CAC would normally deduct bandwidth twice (once on each cluster). To prevent this, inter-cluster trunks are assigned the special Shadow Location. The Shadow Location signals to CUCM that the trunk acts as a transit connection, directing CUCM to deduct bandwidth directly between the true originating and terminating locations.
3. Location Pairs and Link Weights: Administrators define inter-location links between remote locations and regional hubs, assigning an administrative Link Weight (cost) to each link. CUCM computes the effective path, the path with the lowest cumulative weight, between any two locations.
4. Single Effective Path: CUCM deducts bandwidth on every link of the effective path only; it does not try a higher-weight alternate path. If any link on that path lacks bandwidth, the call is rejected or rerouted by AAR.

---

4. Automated Alternate Routing (AAR)

When WAN bandwidth across an inter-location link is completely exhausted, standard CAC behavior drops the call attempt and returns a fast-busy (reorder) tone or a network congestion announcement. In commercial enterprise environments, dropping calls creates poor user experiences. Automated Alternate Routing (AAR) intercepts CAC-rejected calls and transparently reroutes them across the Public Switched Telephone Network (PSTN).

AAR Operational Workflow

1. User A at Branch 1 dials the 4-digit internal extension of User B at Branch 2 (4500).
2. CUCM checks the Locations table for Branch 1 and Branch 2. The inter-location audio pool is at 0 kbps available.
3. CAC blocks the call across the IP WAN.
4. Rather than terminating the call, CUCM checks if User A's device and line have an AAR Calling Search Space (AAR CSS) and an AAR Group assigned.
5. If configured, CUCM dynamically modifies the dialed extension (4500) into a full E.164 PSTN dialable number (e.g., 912145554500) using the destination's external phone number mask and the AAR Group's prefix digits.
6. The call is routed out the local PSTN voice gateway (ISDN PRI or local SIP carrier).
7. The destination gateway at Branch 2 receives the incoming PSTN call, matches the incoming dial peer, and rings User B's telephone.

Core AAR Configuration Elements

+--------------------------------------------------------------------------------+
|                            AAR CONFIGURATION STRUCTURE                         |
|                                                                                |
|  1. Directory Number (DN):                                                     |
|     - Directory Number: 4500                                                   |
|     - External Phone Number Mask: 214555XXXX                                   |
|     - AAR Destination Mask: 2145554500 (or uses External Mask)                 |
|                                                                                |
|  2. AAR Group:                                                                 |
|     - Prefix Digits: 91                                                        |
|     - Assigned to Device Pool or Device/Line                                   |
|                                                                                |
|  3. AAR Calling Search Space (AAR CSS):                                        |
|     - Contains: [PSTN_Egress_PT]                                               |
|     - Only active during AAR rerouting; ignored during standard on-net dialing |
|                                                                                |
|  4. Service Parameter:                                                         |
|     - Automated Alternate Routing Enable: True                                 |
|     - Default: False (must be enabled before AAR works)                        |
+--------------------------------------------------------------------------------+

• AAR Group: Specifies the external prefix digits required to seize a PSTN trunk (e.g., 9 or 91 in North America, 0 in Europe). AAR groups can be assigned at the Device Pool level or directly on the device/line.
• AAR Calling Search Space (AAR CSS): A specialized Calling Search Space containing partitions that access PSTN route patterns or route lists. Under normal calling conditions, the standard Line/Device CSS is used; the AAR CSS is activated only when CAC rejects the primary IP path.
• AAR Destination Mask / External Phone Number Mask: Formats the internal extension into an expanded PSTN string. For example, if the dialed extension is 4500 and the mask is 214555XXXX, CUCM generates 2145554500. Combined with the AAR Group prefix 91, the egress digit sequence becomes 912145554500.
• AAR Voice Mail: Selecting the line's AAR Voice Mail option sends calls that fail CAC to the called user's voicemail instead of rerouting them over the PSTN.

---

5. Diagnostic Metrics & Troubleshooting Counters

Administrators monitor CAC and AAR performance through the Cisco Real-Time Monitoring Tool (RTMT) using dedicated CallManager service counters: