By the end of 2020, video will consume 95 percent of the Internet and business networks, with a greater percentage being real-time or near-real time video. Enterprise’s, cloud and network service providers, who struggled with Voice over IP, will find video to be an even greater challenge. Organizations looking for new SDN architectures to solve their networking challenges will find that SDN does not address the underlying challenges of running video over IP networks.
In Cisco’s annual forecast, they say that by 2020, 82% of all IP traffic (both business and consumer) will be IP video. This is a conservative estimate based on historical and new trends, and does not take into account the magnitude of some of the disruptive factors that are occurring such as:
- Augmented & Virtual Reality – Instead of capturing a video and posting it on a site such as YouTube, more users are streaming real-time video such as Facebook Live and getting video feeds to provide information around them. Collaboration, between users, co-workers, and communities that is video enabled, is becoming more common and starting to blend into our everyday interactions.
- Embedded video analytics – The ability to identify in real-time who and what is occurring in a video is a very exciting development with far reaching impacts. For instance, manufactures can watch their employees assemble equipment, notify them when a mistake is made, and thus further take out defects in the assembly process. This additional intelligence will drive greater usage and value from video.
- Internet of Things (IoT) – Gartner recently increased its forecast of the number of IoT devices by 2020, from 20B to 30B, up from 6.5B today. Video is a key component to an IoT strategy that digital businesses are just now realizing. When event occurs, such as someone ringing your home doorbell while you are at work, you can click to see who it is and to interact with them, then take action such as remotely unlocking the door.
Video empowers users, businesses, and governments to see everything and be virtually everywhere, in real-time. (no comments in this blog on the social implications of this) We move from an age of any device, any user, anywhere, some of the time; to the future of everyone, everything, everywhere, and all the time, in-real time. Our IP networks must rise to supporting all this video traffic, and ensure great performance and security.
Regardless if video is 82% or 95% of the network traffic by 2020, it will be the “killer” application for all IP networks. Real time video challenges existing and future IP networks for the following reasons:
- Consumes Lots of Bandwidth – As anyone who has every run over on their mobile data plan knows, video consumes a lot of bandwidth. As displays get bigger and richer, driving higher resolution video streams such as 4K video, a one hour video session grows from 1GB/hr to 120GB/hr. Plus the WebRTC standard is enabling real-time voice & video to be embedded in any application.
- Variable & Bursty – Unlike most applications, including voice, which have a predictable amount of bandwidth required per session and known limits of spikes, video sessions vary widely on the amount of bandwidth required and spikes can be 10x the base line usage. Modern video codecs are adaptive and will add Forward Error Correction (FEC) when congestion occurs, so a normal 1Mbps video stream that spikes to 3Mbps when there is a lot of movement, will spike up to 10Mbps under congestion, before it backs off. SD-WANs that also provide FEC further exacerbate this problem.
- Dynamic – Networks are static by design and use things like Call Admission Control (CAC) to limit the number of voice and video sessions. This static configuration works marginally for voice and not at all for video because of the above. A fundamental problem with IMS and SDN architectures is that CAC is applied to the session at the time of setup. Once the session is underway, all sessions are treated equally by the network with the same QoS policy since the type of traffic has the same DSCP marking. Not all sessions are equal, which was discussed here back in 2013.
- No Stateful Awareness – Network routing is the only service in the network that is not statefully session aware. Firewalls, load balancers, and WAN optimizers are statefully session aware and thus can control, manage, and report on sessions. While the SDN architecture does separate the control and data planes, it does not do anything to address the lack of session awareness in routing. Without this, a router cannot manage many concurrent video streams effectively such as 20 concurrent Webex sessions.
- Lack of Seamless Roaming – Since video will be on all the time on our mobile devices, we will need the video session to persist across many networks. Today, this process requires changing IP addresses and is very slow and clunky at best, and there is not a global network session identifier to be able to control, log, and report on the session as it roams.
- Low Security Controls – Unlike web or email traffic that use proxies, or VoIP traffic that uses Session Border Controllers, security for video is very weak. The video proxy type devices are proprietary by manufacture such as Edge for Microsoft and Expressway for Cisco. Video runs on top of UDP and uses a wide range of ports, and WebRTC as a new standard is still maturing in areas like security. A man in the middle attack on a WebRTC stream can easily occur, if the video session is not locked down.
Most network operators and enterprises are betting on SDN and NFV to solve today’s IP networking problems. The challenge is that they will not solve tomorrow’s primarily video IP networking problems.
So, as you look at SDN vendors, ask them how they handle lots of video traffic. QoS and CAC are not the answer, so the video based IP networks of the future will need to be architected differently than today’s SDN architectures.