While certain live events have been cancelled amidst the COVID-19 pandemic, many of them have been replaced with virtual counterparts that have taken place via videoconference. Some of these events have had hundreds or even thousands of participants. With such numbers, how do videoconferencing systems cope with the traffic load and processing power required?
In this article, we take a look at the concepts involved with the scaling of videoconferencing for large numbers of participants. We also examine some best practices to keep in mind that will help ensure the quality of the end user experience.
Challenges with scaling
There are two primary challenges involved with hosting an extensively large online videoconference event. One of these is technological, while the other is behavioral.
Videoconferencing is one of the most resource-intensive processes on a network. Both network and system resources are used extensively, especially as the number of participants increases.
Imagine you are hosting a high-definition videoconference with three participants. HD video streams require about 4 to 5 Mbps, which means that each endpoint must send its HD video stream to every other endpoint and each endpoint must also receive every other endpoint’s video stream. You can immediately see how the increase in the number of participants exponentially increases the demand on the network.
Similarly, each endpoint must take the video feed of every other endpoint and process it in such a way so that it can be suitably displayed on their screen. Although the resulting output on the screen of each individual participant will not be in HD, the video stream received is in HD and must be processed as such.
Luckily, much of these fundamental challenges are resolved with the use of a Multipoint Control Unit (MCU), a specialized device that offloads video processing from the endpoints and alleviates network traffic to a great degree.
An MCU is fine for a few dozen or even up to 100 users, but what if you need more capacity?
Calculating required resources for a videoconference
When estimating the bandwidths required for a particular videoconference, the most important values of bandwidth to keep in mind are the following:
- Endpoint outbound (Eo) – This is the bandwidth required for a videoconferencing endpoint to send out its video feed to the MCU. This depends on the quality of the video being sent.
- Endpoint inbound (Ei) – This is the bandwidth required for a videoconferencing endpoint to receive the composite video feed created by the MCU and sent out to all clients. Assuming all endpoints are using the same quality, this is typically about 3 to 4 times the Eo bandwidth, depending on the MCU.
- MCU inbound (Mi) – This is the bandwidth required by the MCU to receive all incoming video feeds from all participants and is equal to the sum of the Eos of all participating endpoints.
- MCU outbound (Mo) – This is the bandwidth required by the MCU to send out its composite video feed to all participants and is equal to the sum of the Eis of all participating endpoints.
Note: the abbreviations used here are not industry standard, but rather are employed to more easily display the equations below. For more information about how MCUs work, refer to our recent article.
Let’s take a scenario where there are 30 participants in a videoconference. All endpoints are using HD video. As a result, we have the following values:
- Eo = 4 Mbps for each endpoint
- Ei = 4 Mbps * 3.5 = 14 Mbps
- Mi = 30 * 4 Mbps = 120 Mbps
- Mo = 14 Mbps * 30 = 420 Mbps
Note here that the Ei is assumed to be 3.5 times the outbound bandwidth. This value depends upon the MCU being used, but is typically between 3 and 4 times the Eo.
The above calculation gives an indication of how an increase in the number of participants can quickly reach the limitations of today’s typical networks, not to mention the processing power required from the MCU. This shows the technological challenges associated with extensively large videoconferences.
Behavioral videoconference elements
Most of us have participated in a small-scale videoconference with up to a few dozen participants. You may have used a cloud-based system like Yealink Meetings, which leverages laptop, desktop, or smartphone software, or a physical videoconference room device such as the Yealink VC800. Regardless of the type of system that you may have used, there are several things that come into play when small-scale videoconferencing is employed, which can affect large-scale videoconferences.
Equal participation – In general, all participants have the opportunity for an equal level of participation. This means that all users can be both seen and heard by everyone else. For conferences with more than five or six participants, it may be necessary to designate a moderator who coordinates this procedure by enabling particular users to speak, but the equal opportunity for active participation still exists.
Etiquette – Proper videoconferencing etiquette requires that you enable your camera for the duration of your participation. Even if you’re not speaking, it is polite to let the active speakers see that you are really there and attentive, rather than them seeing a black box with your name on it. It’s what they would see if you were in a room sitting around the same table.
As you can see, even with the use of an MCU, both networking and system resources can be taxed to their maximum capacity, even for videoconferences of several dozen participants. Similarly, the behavioral limitations of equal participation and etiquette increase the demand upon videoconferencing as the number of participants grow.
Best practices for large videoconferences
So how do we deal with hundreds or even thousands of participants? Some principles to keep in mind are the following:
Unequal participation – Separate participants into two categories: speakers and attendees. Speakers can use full bandwidth requirements and capabilities, while attendees can simply be receivers of the video stream. In this way, attendees don’t actually have to send their video feeds, vastly reducing network requirements and MCU resources. Attendees are almost like receivers of a streaming service rather than a videoconference, but can still participate in other videoconference applications such as chat discussions and polls.
Interconnection with streaming services – Many videoconferencing systems allow a videoconference to be streamed to an external service such as YouTube. This way, you can have an additional category of participant, one that simply follows as they would any other live event on that platform.
Use cloud-based services – For large videoconferences, where equal participation is required from all users, it is always preferable to use cloud-based services rather than employing in-house infrastructure. This is due to the great cost that in-house services will require to support such a large conference. And because cloud-based services are shared, they deliver a greater level of service (network and resource capacity) for a more reasonable and accessible price.
Large videoconferences of all types of events are slowly but surely becoming more commonplace. Not only that, but videoconferencing technology is providing people from all over the world with greater accessibility to these events. Here we have listed some best practices for overcoming the challenges associated with a large number of participants and ensuring a successful videoconference.
You may also like: