Adaptive Mode Mechanism Overview for IPv6

You can configure the auto-sense mechanism in the multicast bandwidth using the set admission-bandwidth adaptive command, set qos-bandwidth adaptive command, or both. For example:

host1(config)#route-map mcast-bandwidths permit 10 host1(config-route-map)#match ipv6 address sdtv host1(config-route-map)#set admission-bandwidth adaptive host1(config-route-map)#set qos-bandwidth adaptive host1(config-route-map)#end

In this example, any stream with an (S,G) that matches the sdtv access list performs adaptive bandwidth detection for admission control and QoS adjustment.

A rate measurement mechanism runs on the ingress line card that polls the forwarding controller (FC) to obtain statistics for each mroute. This mechanism then reports the rate measurement to the switch route processor (SRP) to update the bandwidth map. By computing the average bandwidth over a relatively short sampling period (T1; 5 seconds), the measurement approximates the peak bandwidth of the multicast stream.

As an example, assume that a new mroute (S1, G1) is added to the interface controller (IC) at time t0.

Figure 14: Example of Adaptive IPv6 Multicast Bandwidth Detection

Example of Adaptive IPv6 Multicast Bandwidth Detection

To calculate the measured bandwidth of a stream, the router uses the following equation:

R = (Nt+5 – N t) / 5


R = Calculated bandwidth of the stream during each sampling interval

N t = Bytes measured at the start of each sampling period (t seconds)

Nt+5 = Bytes measured at the end of each sampling period (t+5 seconds)

Note: When the mroute is first installed in the FC (at t = 0), R0 is undetermined. For multicast admission control no joins are admitted until the first bandwidth measurement is computed (that is, for admission control, R0 is considered to be infinite). Similarly, no QoS adjustment occurs until the first bandwidth measurement is computed (that is, for QoS adjustment, R0 is considered to be zero [0]).

Using the earlier graph as a reference, the first bandwidth rate (R10) and at time t5 (N5) and the bytes received values are subtracted and divided by the time period T1 to yield the average rate. This process is repeated every sampling interval, T2, to yield rates R1, R2, R3, and so on.

The first two sampling interval calculations would look like the following:

R1 = (N5 - N0)/5

R2 = (N#+5 - N#)/5

The router maintains a history of bandwidth measurements (H) for each mroute, up to a maximum of M measurements. The actual rate, R, reported to the SRP is the maximum rate measured in those H samples.

In order to minimize the IC to SRP traffic generated by the rate measurements, the IC reports a bandwidth change only when a newly computed rate (R#) differs from the current rate by a specified threshold. When Rs is computed at time t = 5 seconds, R is set to R1. A rate update occurs whenever a newly calculated rate (R) differs from R1 by at least a threshold value (specified as a percentage, P) of the measured peak bandwidth. This calculation would look like the following:

R = Rt, if and only if the absolute value of (R - Rt) > P * R.

The values assigned to variables associated with this algorithm are as shown in Table 50.

Table 50: Adaptive Mode Algorithm Values








Sampling period; the time in which a sample is taken




Sampling interval; zero (0) seconds indicates continuous sampling




Number of history samples over which to compute measurement




Maximum number of samples maintained in history




Threshold value; percent difference by which a newly calculated rate must differ from the measured peak bandwidth before a rate update occurs

Related Documentation