Providing sufficient video conferencing bandwidth to enable high-quality services while also supporting mission critical applications requires careful network assessment, design and management.
There are two problems to solve with regards to video conferencing bandwidth. One problem is ensuring that there is sufficient bandwidth to support legitimate video conferencing usage with high-quality service. The second problem is ensuring video conferencing bandwidth requirements do not impact other mission-critical applications sharing the network. Let’s take a look at how to understand and manage video conferencing bandwidth demands.
Controlling video conferencing bandwidth demands with QoS
The first rule of good video in the enterprise is to properly implement Quality of Service (QoS). This technology will give the right priority to video and audio to ensure they are transported with good quality across the network. But it also means that video can be assigned to its own class, and the bandwidth of that class can be managed.
This critical step allows the network team to determine how much video conferencing bandwidth is allowed on each key WAN link of the enterprise. I recommend creating a spreadsheet that calculates the needed bandwidth for each link, and keeping this document as a specification, agreed-upon by the video conferencing and network teams, on video conferencing bandwidth demands. The network should be configured to drop packets if the video bandwidth exceeds the designed video-class allocation on any WAN link.
Downscaling video conferencing bandwidth demand
Most video conferencing vendors have a feature where the endpoint will step down its bandwidth demand if it is experiencing packet loss on the call. The theory here is that packet loss is caused by congestion, and stepping down the demand will reduce or eliminate the congestion and provide a better video call. I think this is a troubled approach for a number of reasons.
Firstly, network packet loss may not be congestion related, or may be dominated by streams that will just take up the bandwidth if the video backs off. Secondly, there is a tricky timing problem here where congestion may be short or long-lived, and the timing of how video units back off may not match the behavior of the congestion very well. This is a control-loop problem (for those of you who studied classical system dynamics). Lastly, the quality of the video call is degraded as the bandwidth is dropped, so now you may be creating a situation where many users are getting poor quality calls. Save this feature for emergencies, like when a link is down. Design the system to support high quality calls.
Configuring call managers for quality call control
Video conferencing systems are managed by a call manager or gatekeeper that has control over if and how calls are placed. These systems can be configured to force video endpoints to use a specific call bandwidth, thus preventing a few users from dominating the available bandwidth with high bandwidth calls. Determine the bandwidth that provides sufficient quality for your business application and then configure the call manager to enforce this rule.
Call managers can also track the bandwidth used on each link of the network. Typically call managers are programmed to understand a basic WAN network diagram, and bandwidth is assigned per WAN link. This pool of bandwidth is then allocated to calls as they are set up, and de-allocated when they finish. This process allows the call manager to determine if the next requesting call will overload one of the WAN links or not. Calls that overload the network can then be denied, rerouted or allowed in only at a lower bandwidth that does not exceed the link capacity.
Creating a lower QoS class for desktop video conferencing
Recently I have worked with several enterprises that are deploying large numbers of desktop video conferencing systems, and they are rightfully concerned about the bandwidth these endpoints will create. Desktop video conferencing demand can grow very rapidly because the deployment of additional endpoints is so easy. Some of these enterprises are choosing to create a separate QoS class for desktop video conferencing, to separate its demand from the demand of the room-based and telepresence systems in the same company. This separate class has its own bandwidth allocation which, if over-utilized, will not impact the quality of the existing room-based conferencing or telepresence infrastructure.
Monitoring bandwidth utilization
Critical to managing this growing demand for video conferencing bandwidth are tools that will let the IT team see how much bandwidth is actually being used, the timing of peak demand, how close the actual demand is to network limits and the trend over time. There are a couple of ways to monitor this traffic, and I usually recommend that enterprises implement both.
First, use NetFlow or sFlow to monitor the actual use of the video conferencing QoS class on the key WAN links of the network. NetFlow (Cisco) or SFlow (others) samples the traffic flowing through WAN links by counting packets as they pass through the key routers. A number of network management tools can act as a front end to these technologies to create the appropriate graphs and reports.
Second, consider a tool that can analyze the call data records (CDRs) created by the video conferencing infrastructure, such as the SMART tool from DesigNET. The CDRs contain information about each call -- including the endpoints involved and the call rate -- and may additionally have information about call quality, call consistency and network behavior. Bandwidth utilization over time is presented by aggregating all the calls across each key WAN link to see the average and peak demand for that link.
Managing video conferencing bandwidth demands in summary
My approach is to design the network to support the expected demand, using the demand spreadsheet. Establish an agreement between the video and network teams using this document. Design the network to support the expected demand. Then ensure that the video conferencing bandwidth demand never exceeds the designed amounts through bandwidth management (call admission control). Lastly, monitor actual utilization, and renegotiate the bandwidth design as the use increases and the company realizes the benefits of video conferencing.
About the author:
John R. Bartlett is a principal consultant at NetForecast, where he focuses on network support for voice and video conferencing. NetForecast provides consulting to enterprises and networking equipment vendors on application performance issues and convergence of voice and video conferencing on the IP network. John has 32 years of experience in the semiconductor, computer and telecommunications, and has been consulting since 1996. John can be reached at firstname.lastname@example.org.
This was first published in December 2010