Medianet Reference Guide - Cisco

Chapter 6
Medianet Management and Visibility Design Considerations
Cisco Network Analysis Module
definition video for several reasons. First, video flows are sent as groups of packets every frame interval.
These groups of packets can be bunched-up at the beginning of the frame interval, or spread evenly
across the frame interval, depending on how the video application (that is, the transmitting codec) is
implemented. Also, each packet within a single frame can vary in size. Operations such as the UDP jitter
IPSLA operation transmit fixed-sized packets at regular intervals, similarly to VoIP. Second, high
definition video frames often consist of more than ten packets per frame, meaning that the interval
between the individual packets sent within a single video frame can vary from less than one millisecond
to several milliseconds. Observations on a highly loaded Cisco Catalyst 6500 with a Sup-32 processor
have shown that individual UDP jitter IPSLA operations can generate packets with intervals of
4 milliseconds or greater with large packet payload sizes. Smaller platforms such as the Cisco 2800
Series ISR may be capable of generating packets with intervals of only 8–12 milliseconds or greater,
depending on the loading of the platform.
Note
Although some platforms allow configuration of intervals down to one millisecond, the design engineer
may find it necessary to capture a data trace of the IPSLA probes to determine the actual frame rate
generated by the IPSLA sender. Partly because the loading on the CPU affects the rate at which IPSLA
probes are generated, pre-assessments services of deployments such as Cisco TelePresence are often
performed with dedicated ISR router platforms. Likewise, some organizations deploy router platforms
at campus and branch locations dedicated for IPSLA functions.
Crafting one or more UDP jitter IPSLA operations that accurately replicate the size of the individual
packets sent, the interval between individual packets sent, and the frame-based nature of video can be
challenging. These attributes are important to factor in because network parameters such as jitter and
packet loss are often largely dependent on the queue depths and buffer sizes of networking gear along
the path between the endpoints. Sending a smooth flow of evenly spaced packets, or larger packets less
frequently, may result in significantly different results than the actual video flows themselves.
As an example, to accurately pre-assess the ability of the network to handle a flow such as a TelePresence
endpoint, you must craft a sequence of packets that accurately simulates the endpoint. Figure 6-17 shows
a close-up of the graph from a data capture of the video stream from a Cisco TelePresence CTS-1000
running Cisco TelePresence System version 1.6 software.
OL-22201-01
Medianet Reference Guide
6-25