White Paper

White Paper

A typical enterprise customer has various organizations, such as engineering, human resources,

marketing, sales, and support. The traffic patterns are different in these organizations. You can

expect more calls for support, sales, and marketing groups, and the switches that house these

users will have a higher busy hour call attempt (BHCA) value. In a call center environment, you can

expect high call volume and, therefore, high BHCA, which will be the deciding factor for the number

of sensors required.

The number of simultaneous calls is tied to the previous argument about the BHCA value. It is

important, therefore, to understand the BHCA value and average call hold (ACH) time to determine

the number of sensors required. Keep in mind the limit on the number of RTP streams supported

by each sensor (100) and the number of simultaneous calls generated by the phones connected to

the switch where the sensor is deployed.

You could have a situation in which the number of RTP streams exceeds 100, in which case you

can add sensors or configure the SPAN source port in such a way that you monitor only selected

phones on the switch. If the SPAN destination port sees more than 100 RTP streams, the sensor

goes into sampling mode and the MOS value reported is diluted.

Note: 1040 bandwidth requirements are very little. Given that syslog is around 60 bytes per

stream, so max syslog each minute would be 6000 bytes when reporting 100 streams each minute.

Skinny keep alives are every 30 seconds and a heartbeat message only has about 12 bytes of TCP

payload for each KA, and KA-ACK.

Cisco 1040 Sensor in Sampling Mode

A Cisco 1040 Sensor can monitor 100 RTP streams. If it is deployed on a switch that has more

than 100 RTP streams, it performs sampling, in which case some of the RTP streams are not

considered for MOS value generation. You must avoid this situation at all times. In sampling mode,

the MOS value reported is diluted because some of the RTP streams are not considered. The

sensor monitors the RTP stream and collects the information necessary to compute the MOS

value. This information is stored in a buffer, from where the computation process picks data to

compute the MOS value. If the packets arrive at a faster rate than the rate at which the buffer is

emptied, part of the RTP stream is dropped before the sensor starts the collection process. Keep in

mind that CPU resources are constantly utilized; hence, it is not just the buffer that becomes a

bottleneck when the sensor is overwhelmed with more RTP streams: the CPU also falls short in

serving the different processes. The MOS value reported by the sensor gets worse as the number

of simultaneous RTP streams increases beyond 100. It is important to plan ahead and optimize the

SPAN port configuration in these scenarios.

Cisco 1040 Sensors in the Branch

In a branch office, the density of IP phones is less when compared with the density seen in a

campus. Typically, the branch office contains fixed-configuration switches, and the number of

simultaneous calls is lower.

In a fairly large branch, it is common to see multiple fixed-configuration switches stacked to provide

more density and avoid the need to run gigabit uplink to aggregate switches and routers. The Cisco

1040 Sensor fits into this model the same as with any other switch. It utilizes the SPAN port to

monitor the RTP stream.