The development of new Internet Protocol (IP)-packet based Triple Play services (voice, video, data) delivered over the telco Access Network places new demands on network designs. Each of the services has their own specific quality of service (QoS) requirements, and all have varying bandwidth demands that are dynamic in nature, and, as a result, require prioritized treatment of the 3 services. In the Access Network, where bandwidth (BW) is often limited due to Digital Subscriber Loop (DSL) links, for example, BW allocation and control is critical. The mix of services supported must also match the customer subscriptions and thus expectations: the number of high definition (HD) video streams, standard definition (SD) video streams, and voice calls supported simultaneously, and the amount of BW provided for data service.

Many mechanisms exist for delivering the proper QoS for each service application. Several network design concepts are typically used to identify the data flows related to the 3 service types; and based on that, enable network equipments to treat the flows differently when resources become limited. The interaction between competing application bandwidth demands may affect the individual QoS. Testing individual applications in a sterile environment will not reveal problems that manifest in the presence of mixed application traffic flows, thus requiring a new approach to analyze these interactions: a Class of Service (CoS) approach.

CoS is a new test process that refers to the treatment of different applications with associated requirements for how data flows are handled in the network. A CoS network design must include strategies for:

1. Prioritizing traffic dealing with congestion caused by peak BW demands, which can result in data loss.
2. Admission control that can manage service conflicts.

Strategies to do this include:

• Resource and admission control (RAC).
• Virtual local area network (VLAN) application segregation.
• Multiprotocol Label Switching (MPLS) guaranteed BW.
• MPLS fast reroute.
• Policing and marking on ingress routers.
• Differentiated queuing and dropping on core links.
• Shaping (or policing, based on line speeds and hardware deployed in the network) and differentiated queuing on egress links.

Regardless of the approach a given network may utilize, validation of CoS performance remains critical, especially in the Access Network where BW is often limited. The interaction between competing application BW demands may affect the individual QoSs. Testing individual applications in a sterile environment will not reveal problems that manifest in the presence of mixed application traffic flows.

The Reality
The true test of a network and how the CoS mechanisms are working is realized when sequencing multiple applications online and then gaining an understanding of the interaction between the different services. In the example shown in Figure 1, the maximum DSL BW is 25 Mbps. Therefore when considering the various encapsulations, the actual useable BW for different services is about 4 percent less, in this case 24 Mbps.

Figure 1. Sequencing multiple applications online.

As the graph in Figure 2 shows, stream 1 (red) is a Standard Definition (SD) video stream with a variable bit rate averaging about 3.75 Mbps. Bringing Stream 1 online first is followed by the sequencing of the data stream (light blue) online. This stream shows the BW varying during a window size negotiation before peaking at about 12.25 Mbps, and shortly thereafter adding a third stream (dark blue) to the mix.

Figure 2. All bandwidth allocated.

This is a High Definition (HD) video stream with a variable bit rate and an average BW of about 8.75 Mbps. At this point, the total average BW is about 24 Mbps, essentially allocating all of the available BW in this particular example. (At the cursor position, the color-coded display shows the exact BWs on the data: Video 1, 3.32 M, Video 2, 9.04 M and Data, 11.5 M with a total of 23.86 M.)

When sequencing a third video stream online, an HD stream (green) could be running at 8.75 Mbps. Therefore, the total BW demand exceeds the 24 Mbps maximum.

In reaction to this BW demand, the network compensates by simply modifying the BW for all streams so that the total BW demand is 24 Mbps. Video 1 is squeezed to 2.5 M, Video 2 to 6 M, Data to about 10 M, and Video 3 to 6 M, for a total of 24 Mbps. As a result, it destroys the QoS for the video services. Massive packet loss occurs on all 3 video streams, and the quality of the video and audio becomes unacceptable to the user. This is an example of a network where the Admission Control and CoS mechanisms are not working or have not been implemented properly.

An alternative test approach may address each application individually. This approach would result in a total BW demand of about 21 Mbps for the video application, (all 3 video streams: 8.75 + 3.75 + 8.75 = 21.25 Mbps) under the 24 Mbps maximum. There would have been no packet loss, leading one to believe the system was operating correctly.

However, with no data application active, the impact of a real-world mixed application environment would remain untested and the data service impact would remain unseen, resulting in customer trouble calls upon establishing the mixed traffic.

Example #2
In another example shown in Figure 3, video service is capped at 15 Mbps leaving BW for voice and data services. The first screen shows 2 video streams active: Video 1 SD stream with a BW of 3.463M, and Video 2 HD stream with a BW of 8.235M, for a total of 11.737M as shown by the white line graph.

Figure 3. Mixed applications with 2 (capped) video streams.

Figure 4 clearly shows the effects of adding a third video HD stream. The 15 M limit is exceeded. The network responds by reducing the BW for the 2 HD streams from about 8M each to 6.254M and 5.938M resulting in massive packet loss. The total line graph shows the new combined total BW at 15M. Adding a data flow to this particular test would reveal that the data application would not impact the video flows since data service BW is reserved separately from the video service in this network example.

Figure 4. Limit exceeded by adding third video stream.

In each of these cases, it is difficult to identify the root cause or perform timely trouble resolution without the ability to test the interaction of video, voice, and data services together. The individual application tests cannot duplicate the failure or clearly show the interaction between streams.

With the many different configurations deployed in today's IPTV networks, use of this CoS test concept is important for proper, new service installation testing. Validating that the CoS setup is working correctly is critical. It is also a valuable tool for use in trouble resolution testing.

Never Out-Classed
In summary, CoS testing enables field technicians and engineers to validate that IP services are performing correctly in a real-world mix of applications. In the Access Network, where BW is often limited, CoS analysis may reveal interaction problems or BW-limitation problems that would otherwise remain unidentified.

Using CoS testing, all 3 applications can run simultaneously to validate whether multiple streams can run at once and to ensure that the network CoS mechanisms are working correctly. Adding new streams lets technicians see if a negative impact occurs on the existing streams.

 About the Author
John Williams, Director, Emerging Markets, has been with JDSU for more than 15 years and has served in senior-level corporate and technical management positions. Currently he manages research for all emerging technologies related to telco Access Networks and determines what new test opportunities and potential alliances are developing. Mr. Williams oversees JDSU's IP video development and continues to help define the IP video roadmap and manage enhancement releases for JDSU products. For more information visit www.jdsu.com.

JDSU offers the CoS Test Suite via a patented software option available for the HST-3000 Handheld Services Tester and T-BERD®/MTS-4000 Multiple Services Test Platform. For more information, visit www.jdsu.com.

What’s your take on this subject? Leave a comment and get the conversation going.