See What You’ve Been Missing:
By Oleg Berzin, CCIE 3955Doing Business in High Definition
Introduction
As many may agree 2008 was a challenging year – the economic slowdown has been taking its toll on the business side of the technical world. However the technology side of the technical word unfolds fairly independently from the fluctuations of the economic situation. New technologies and applications emerge and the evolutionary transformation of networking goes on. One very notable contributor to this transformation in 2008 has been HiDef business Videoconferencing or as it’s called TelePresence.
Most of us have experienced the benefits of High Definition (HD) television in both the quality and the form factor. No more bulky entertainment centers, no more dusty cable tangles – just the screen and a great picture. In a way clearing the space also cleared the path to a new level of perception of the entertainment content. Similarly, the expansion of HD technologies brings changes not only to the way we view entertainment as the consumers of HD products and services but also to the way we conduct business and enhance our productivity.
TelePresence systems are the next generation of the video conferencing products. Applying the family room analogy, by making use of large flat screens, High Definition video and High Fidelity audio TelePresence minimizes the presence of equipment and maximize the presence of content. And in particular, for the TelePresence application, the content are the people on the far end of the connection, that appear in life-size, high-resolution images with natural eye contact and voice at the other end of the meeting table.
Needless to say there is no magic behind the new technology. As it turns out there is a price to pay for TelePresence – it creates a new class of applications and the corresponding traffic properties – High Volume, Variable Rate, Real Time Interactive applications. The secret to TelePresence is not only in the system itself but also in the network on which it is carried.
As engineers we understand the technical aspects behind a solution as well as the importance of the user expectations. Since the TelePresence systems are marketed as the next generation true HD business video conferencing and are sold with a premium price tag the users expect zero visually noticeable picture impairments. This is similar to the consumer expectations for the HD products and services. On the other hand the main contributor to the possible impairments is the network connection between the TelePresence room systems. Thus it logically follows that the network used to carry the TelePresence traffic must be designed, engineered and provisioned to ensure a very low probability of visually noticeable picture impairments.
System Tolerances and Network Requirements
The combined High Definition Video and Hi Fidelity Audio traffic stream generated by the Cisco TelePresence System (CTS) represents an unconventional mix of an interactive application and variable rate, burst and packet size streaming video associated with specific requirements for packet loss, jitter and latency. The effects of violating the TelePresence system tolerances for packet loss, jitter and latency may lead to dramatic loss of picture quality and the TelePresence experience.
TelePresence System tolerances define the range of parameters within which the system is expected to operate normally on an end-to-end basis. These parameters may be specified in system specific terms and units of quantity and time. In particular, the following System Tolerances and Parameters are applicable to the Cisco TelePresence System.
- One Way “Camera-to-Glass” Latency. This is the time (in milliseconds) that it takes a Video/Audio signal generated at the source TelePresence system to appear at the destination TelePresence system. This is also referred to as the “Camera-to-Glass” time.
- One Way Network Delay. This is the Camera-to-Glass latency minus the CODER/DECODER delays on each side.
- Packet Loss Ratio. This is the ratio of packets that were lost during the network transport to the total number of packets sent by the system for the duration of time over which the TelePresence session was active.
- Packet Jitter. This is the variation in One Way Network Delay times experienced by the packets sent by the system over the duration of the session. Specifically, the Packet Jitter is measured on a Packet-to-Packet basis as the variation between one-way Network Delay values for successive packets. The system tolerance for this parameter is expressed as the Maximum Absolute Value of the Packet-to-Packet Jitter in milliseconds (Peak-to-Peak).
- Averaged Information Rate. This is the maximum expected traffic rate that results from the Cisco TelePresence system Video and Audio stream multiplexing process averaged over a one second time interval. This value depends on the TelePresence video resolution and motion handling settings. This parameter however, does not reflect the burstiness of the TelePresence traffic and should be adjusted when the logical port speeds are provisioned in the network.
- Statistical Burst Profile. This is also referred to as microbursting. This is the statistical distribution of the traffic burst sizes and their corresponding durations generated by the system. Although the Averaged Information Rate is a useful parameter it does not reflect the traffic fluctuations related to the video stream carrying the frames resulting from the high motion periods of the session. The burst rates may be significantly higher than the Averaged Information Rate over much smaller periods of time (tens of milliseconds). Since the full statistical burst profile is quite difficult to obtain a burst margin is used to accommodate for traffic fluctuations. The burst margin is added to the averaged information rate and is expressed in kilobits per second.
- Protocol Overhead Margin. This is the amount of additional L2/L3 information required to correctly encapsulate TelePresence bearer traffic. The protocol overhead margin is expressed in kilobits per second and is estimated by calculating the number of additional bytes related to the protocol headers/trailers and the expected packet per second rate of the TelePresence bearer stream.
| System Tolerances / Parameters (CODEC-to-CODEC) | Degree of Loss of Video/Audio Quality if not met |
|---|---|
| One-Way Network Delay: Less then 200 milliseconds |
Moderate |
| Packet Loss Ratio: Less then 0.05% for the session duration |
Very High |
| Packet Jitter: Maximum Packet-to-Packet Jitter Less then 20 milliseconds for the session duration |
Very High |
| One-Second Averaged Information Rate( Example Cisco CTS3000): 15 Mbps |
Very High |
| Burst Profile (Example Cisco CTS3000): 20 Mbps over 25 milliseconds |
Very High |
A Challenge for the Network
As can be seen from the system tolerances, the requirements for transporting TelePresence traffic over a packet network are quite different from other traffic type requirements including VoIP. Although some level of similarity in the types of traffic parameters (Delay, Jitter, Loss) exists between TelePresence and VoIP there are significant differences.
Specifically, the Jitter and Loss requirements for TelePresence are much stricter than for VoIP. VoIP traffic can survive network conditions with packet loss values of 1-2% and packet jitter values of 30-50 milliseconds without a noticeable degradation in speech quality. In addition, VoIP traffic exhibits a constant packet rate with fixed packet sizes (for example a G.711 codec would produce 200-byte IP packets at a rate of 50 packets-per-second). This makes it convenient to provision the network transport for VoIP following conventional voice traffic engineering models where a fixed trunk bandwidth is reserved based on the required bandwidth per call and the expected busy-hour call volume.
With TelePresence in addition to the three main Network Requirements critical to maintaining the integrity of the Video and Audio streams (Delay, Loss and Jitter), the other fundamental traffic control requirements are the Traffic Policing and Burst Control.
Traffic Policing is a rate limiting technique that is based on the metering of the incoming or outgoing traffic over a preset time interval. The metering time interval is specified explicitly or derived from the specified Committed Information Rate (CIR) and the Committed Burst Size (CBS or Bc). The Traffic Policer determines if the traffic is compliant by calculating the number of bits (or bytes) received over the metering interval (traffic burst) and comparing that number with the specified CBS/Bc. The bits/bytes received in excess of the CBS/Bc value are discarded by the Traffic Policer.
Burst Control is associated with Traffic Policing and is a method of specifying the degree to which the incoming or outgoing traffic is allowed to vary its rate within the metering time interval. If the metering time interval is relatively large and the CBS/Bc size is also relatively large the incoming or outgoing traffic has greater flexibility in its rate variations. For example, in the beginning of the metering time interval the traffic source may transmit at a very high rate for a short portion of the overall interval, while for the rest of the metering time the source may send at a very low rate.
One of the important “hidden” challenges for the QoS enabled network with respect to supporting TelePresence traffic is rooted in the fact that the existing Real Time traffic class that is capable of maintaining the integrity of the HD video streams by providing low latencies, packet loss and jitter was designed and engineered specifically to support VoIP. In particular, due to the predictable nature of VoIP traffic, this means that the network elements handling this traffic class are usually provisioned with relatively small Committed Burst Sizes (or Bc values). For TelePresence video streams this has a great implication in that the micro-bursts of video traffic may not be fully absorbed by the network element’s Committed Burst buffers, which leads to packet loss, which in turn causes a dramatic degradation in the HD picture. This is also often evident in Ethernet Access networks where seemingly reasonably sized Ethernet Virtual Circuit rates provisioned over a 100 Mbps physical interfaces are not capable of absorbing the video traffic bursts due to small Committed Burst Sizes.
A Challenge for the Designer
As the designers of end-to-end solutions we have to understand all of the variables and their interplay in order to deliver a reliable system. In the case of TelePresence, most popular designs involve transcontinental network connections, understanding and interacting correctly with the service provider’s QoS and traffic management policies, CPE router and LAN based QoS for the signaling and bearer traffic, protocol overhead implications and survivability.
Needless to say, end-to-end testing and verification are imperative for ensuring that all these variables are accounted for and that the overall design satisfies the system tolerances. It is highly beneficial to be able to test the TelePresence behavior in a controlled lab environment so that when it comes to the field deployments the unpleasant surprises are minimized.
Testing and Verification
The overall objective of the verification and testing of the TelePresence system is to establish a composite network impairment profile within which the TelePresence system is expected to be able to maintain its end-to-end integrity. The composite impairment profile is defined by a three-component parameter vector {Delay, Loss, Jitter}. The TelePresence system is expected to operate as designed if the network impairments do not exceed the composite impairment profile and the network is providing a minimum service rate (including the required margins) for a given resolution level.
Figure 1. Composite Impairment Profile
The overall setup of the test and verification environment is shown in Fig. 2.
Figure 2. TelePresence Test and Verification Environment.
TelePresence Systems
The two main examples of the commercially available TelePresence systems are Cisco TelePresence (CTS) and Polycom RealPresence (RPX). Both systems are targeted to deliver an in-person meeting experience while communicating over a network. The systems consist of a TV studio level conferencing room with strictly specified color, lighting and noise requirements, and employ the use of high definition cameras, high fidelity microphones and speakers as well as the large high definition monitors.
What’s Ahead
As with any other technology evolution of the TelePresence systems and services will most likely proceed in two major ways: existing systems will be enhanced and become an integral part of the enterprise business environment, and based on the existing technology new and more exciting systems will emerge.
Clearly the commoditization of existing technology requires a streamlined network design process in which all of the variables related to system parameters, tolerances and quality of service are well understood and the design itself is transformed from being “more of an art” to being “more of a steady process”. As for the future, we have already seen a glimpse of it from Cisco – a 3D HiDef Holographic TelePresence system. And what’s ahead for us – the networking professionals? It’s the new challenges and problems to be solved.
Oleg Berzin is a Principal Member of Technical Staff with Verizon Communications. He has over 14 years of experience in the Telecommunications Industry including Network Architecture, System Analysis and Design, as well as the extensive operational expertise in a wide range of telecommunication and networking protocols and products. His area of expertise is designing, evaluating and implementing complex multi-technology, multi-protocol networks. Oleg is a Triple CCIE – in the Routing & Switching, WAN Switching and Service Provider specializations. Currently Oleg is a doctoral student at Drexel University in Philadelphia, PA. His areas of research include Mobility Management, Quality of Service, Network Architectures and Protocols.
Send Feedback
Return to the top of this article.