Sponsored By

Overlay or Converged Network for Telepresence?Overlay or Converged Network for Telepresence?

The boss as asked you to deploy telepresence and your job is the network. The first decision to make is: Do we implement an overlay network or converge the telepresence traffic on the data network? An overlay network means a new set of LAN and WAN connections dedicated to the support of the telepresence traffic. A converged solution means carrying the telepresence on your current network and just increasing the bandwidth where necessary. Which is the right approach? It depends on your enterprise. There are three key topics driving this decision; let's take a look at them and evaluate the tradeoffs.

John Bartlett

June 16, 2008

4 Min Read
No Jitter logo in a gray background | No Jitter

The boss as asked you to deploy telepresence and your job is the network. The first decision to make is: Do we implement an overlay network or converge the telepresence traffic on the data network? An overlay network means a new set of LAN and WAN connections dedicated to the support of the telepresence traffic. A converged solution means carrying the telepresence on your current network and just increasing the bandwidth where necessary. Which is the right approach? It depends on your enterprise. There are three key topics driving this decision; let's take a look at them and evaluate the tradeoffs.

The boss as asked you to deploy telepresence and your job is the network. The first decision to make is: Do we implement an overlay network or converge the telepresence traffic on the data network? An overlay network means a new set of LAN and WAN connections dedicated to the support of the telepresence traffic. A converged solution means carrying the telepresence on your current network and just increasing the bandwidth where necessary. Which is the right approach? It depends on your enterprise. There are three key topics driving this decision; let's take a look at them and evaluate the tradeoffs.First on the list is Quality of Service (QoS). Telepresence is interactive video conferencing, which is a real-time application. Because real-time traffic is very different than data traffic (see previous post), a well engineered QoS solution is vital. If the current network does not yet implement QoS, there is a lot of work to do to design a QoS strategy, get it implemented in all routers and switches and verify it is operating correctly. The impact of the telepresence application on the existing data applications must be assessed to ensure they will continue to work with adequate performance.

If deploying QoS on the enterprise network will be a big task, consider using the overlay model instead. Because only telepresence traffic is carried on the overlay connections, the implementation of QoS is much simpler. It is still needed, in order to separate signaling and management traffic from video and audio traffic, but it is more forgiving and much easier to get right.

Second on the list is bandwidth. Telepresence is a large consumer of bandwidth. And the bandwidth that it consumes is constant, so no statistical multiplexing is possible. I will spend more time on calculating the bandwidth needed for telepresence in my next post. But the tricky part about bandwidth is determining how much is needed for the existing data applications. Data applications require bursts of bandwidth when users require new data on their screens or when they access a database. If there is insufficient headroom in the network to support those bursts, application performance suffers.

An overlay network solves this problem by isolating the telepresence traffic on its own links. These links can then be utilized to very high levels because the telepresence bandwidth is both predictable and well behaved (not bursty.) It is easier to predict the performance result of an overlay and it eliminates the potential impact on data applications. In some cases, it may also be more cost effective to implement the overlay because of the high utilizations possible on these dedicated links.

The third decision point is the schedule. While both of the issues cited above can be resolved with careful engineering, this work takes time to design, implement and verify. If the telepresence decision came from the top, as it often does, the schedule may also have come from the top and may not provide sufficient time to get convergence done right. For these scenarios the overlay network can again provide relief because it is much simpler to deploy.

Don't despair, the overlay does not have to be the final solution. The overlay network can be merged back into the converged network once the right QoS is deployed and the bandwidth issues are well understood. Long term, the converged network will provide the best cost efficiency. But while the network evolves and the network team learns the details of supporting large volumes of real-time traffic, the overlay network can provide a simpler solution.

About the Author

John Bartlett

John Bartlett is a principal with Bartlett Consulting LLC, where he provides technical, financial, and management leadership for creation or transition of Unified Collaboration (UC) solutions for large enterprises. John discovers the challenges in each enterprise, bringing disparate company teams together to find and execute the best strategy using Agile-based methodology to support quick wins and rapid, flexible change. John offers deep technical support both in collaboration solutions and IP network design for real-time traffic with global enterprises world-wide.

 

John served for 8 years as a Sr. Director in Business Development for Professional & Managed Services at Polycom. In this role he delivered, defined and created collaboration services and worked with enterprises to help them shorten time-to-value, increase the quality and efficiency of their UC collaboration delivery and increase their collaboration ROI.

 

Before joining Polycom, John worked as an independent consultant for 15 years, assessing customer networks for support of video applications and other application performance issues. John engaged with many enterprises and vendors to analyze network performance problems, design network solutions, and support network deployments.

 

John has 37 years of experience in the semiconductor, computer and communications fields in marketing, sales, engineering, manufacturing and consulting roles. He has contributed to microprocessor, computer and network equipment design for over 40 products.