No Jitter is part of the Informa Tech Division of Informa PLC

This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them. Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.

Quality of Service (QoS) Design for Telepresence

QoS is of course required if we expect to get consistent high-quality images and audio. To work correctly, real-time applications like Telepresence need the high priority that QoS provides. Packets need to be delivered with low loss and jitter so that the real-time images and audio can be reproduced as they happen. QoS is required both for an overlay network as well as for a converged network.

In an overlay network, the voice and video streams only compete with the signaling and management traffic that supports telepresence. QoS is still needed, but a simple 2-level strategy is sufficient. This is a key reason to go with an overlay if you have a short schedule and a grumpy boss.

In a converged network, telepresence has to be integrated with the other applications requiring priority service including VoIP, interactive data applications like Citrix, and mission critical Web-based applications. Telepresence needs to be higher in the priority stack than all these applications except the VoIP.

Here is where the details about bandwidth become most important. The QoS we most often use in our networks today is DiffServ, which by its strictest definition is a Class of Service (CoS) technology. CoS will provide a known forwarding behavior to a class of traffic, such as the telepresence, but will not manage its bandwidth. The routers will establish a maximum bandwidth for the telepresence class on each link. If that bandwidth is exceeded, the routers begin to drop packets, and thus impact the quality of the telepresence call.

The bandwidth design must be done carefully as described in an earlier post. If there is insufficient bandwidth for the telepresence, quality will suffer. Likewise if there is insufficient bandwidth for the data applications when the telepresence is running, the data applications will exhibit slow response and users will complain.

What class of service should be used for telepresence? Older literature from Cisco recommends that video conferencing traffic use an Assured Forwarding class (AF41). VoIP uses an Expedited Forwarding (EF) class. The difference between these two is that an assured forwarding class guarantees bandwidth, but does not guarantee that each packet in the AF queue will be forwarded immediately. The router uses a weighted round-robin algorithm to give just enough favor to the AF queue to ensure it is able to transmit at its designed bandwidth maximum. The EF queue, on the other hand, always gets to forward its traffic as the next packet to go. An EF queue ensures the lowest latency and jitter for a real-time application.

More recent Cisco literature and the presentations they have been using at the VoiceCon session on Voice-Ready Data Networks have a different story. Here they show telepresence being integrated into the EF queue with special thresholds to ensure that both the voice and video passing through the queue get EF priority but can still be bandwidth-managed separately. Interesting that video gets a better priority now that Cisco is in the video business.

RFC-4594 is a set of guidelines for sorting out the priority of each application in your QoS strategy. This RFC is the aggregated knowledge of some of the best network minds in the industry who have been thinking about this issue for many years. Understand in detail the QoS support provided by your WAN service provider and what markings they expect. Start with those two reference points and work out the best QoS strategy for your specific deployment and needs.