White Paper

ManualsBrandsCisco ManualsCloud and Systems ManagementCisco Prime Unified Service Monitor 8.7

White Paper

covering segments 1 and 2. The RTP stream coming from Phone B covering segment 4, 3,

and 2 is of importance; the switch on the left must be configured to span the incoming RTP

stream, to span the destination port seen by the left sensor.

With optimal SPAN port configuration, each sensor can monitor 100 calls.

The number of sensors required depends on the busy hour call completion (BHCC) handled by the

switch. For a 10,000-phone network, for example, the cluster could handle 1000 to 4000 call

simultaneously. As the size of the network increases, it becomes more appropriate to take samples

of the calls generated from the cluster and measure voice quality for a subset of these calls. On the

other hand, if the network consists of only about 1000 phones, it is easier to monitor voice quality

on all of the calls; however, the same sampling technique can also be applied to a 1000-phone

network.

The Unified Communications deployment follows one of the following call processing models:

●

Single site with centralized call processing

●

Multiple-site WAN with centralized call processing

●

Multiple-site WAN with distributed call processing

For a single site with centralized call processing, most often a Catalyst

6500 is used in the access

layer/wiring closet to connect the IP phones. It is common to expect about 200 IP phones (4 blades

with 48 ports = approximately 200) on a single Catalyst 6500. If 4 out of 10 phones are active at

any point, a single Cisco 1040 Sensor can be placed on the SPAN port of the Catalyst 6500 to

monitor the active calls. It is possible to deploy multiple sensors on the same switch to address

situations in which the switch is handling high call volume.

It is not necessary to measure voice quality for every call. The general practice is to measure voice

quality for a subset of the calls on a continuous basis and use a tactical approach for

troubleshooting voice quality problems.

Based on this analysis, for a 1000-phone deployment, if you were to sample 30 percent of the

active calls and measure voice quality on a continuous basis, you would need three Cisco 1040

Sensors (30 percent of 1000 is 300, and each sensor can monitor 100 RTP streams; hence, you

would need three sensors). Usually, the sensors are deployed in pairs (one each at the origination

and termination endpoints).

This sampling could be quite aggressive for most common deployments. Based on the

simultaneous calls that are active on a switch, the number of sensors required for voice quality

measurement varies. As the network size increases, the sampling policy can be reduced.

The previous scenario had a dense population of phones on a single switch; another scenario

could have phones scattered among multiple smaller switches. Typically, the number of phones in

a branch is less, and you will see smaller-density switches used in the branch. For this scenario, it

is excessive to have one sensor per switch, especially if the number of RTP streams on these

switches is small (20–40 RTP streams). To address this issue, you can use Remote SPAN

(RSPAN) and combine RTP streams on multiple switches. Another alternative is to use an active

hub and connect the SPAN destination port from multiple switches to the same active hub. With

this alternative, there is potential for Layer 2 loops, and you must evaluate the best option before

embarking on any approach.

In summary, Service Monitor can monitor actual voice calls in real time and provide details of the

parameters that cause voice quality degradation. The combination of Service Monitor and