How to Maximize the Reliability of a Cellular Bonding Solution
Cellmuxes are no longer novelties reserved only for the most extreme shooting conditions. They now enjoy pride of place in every field video producer’s toolbox and play an integral role in electronic newsgathering. At the 2014 World Cup one vendor reported that more than 200 of its units were used each week during peak times, with more than 40TB of data delivered over the course of the entire event.
Of course, no matter how popular cellmuxes and channel-bonding solutions become, they’ll never offer the same reliability as more tried and true backhaul methods. This article will examine ways that you can ensure the greatest uptime and throughput when you’re using these devices in the field.
First, we need to be clear on some definitions: “channel-bonding” refers to cellmux devices (which include video encoding), and not simple channel-bonding link aggregators (which can work to create a single high-capacity Layer 3 network, but with no video intelligence).
While it might seem pedantic to separate different classes of device out like this, it is this reckless misuse of the terms “channel bonding” (or sometimes “3G bonding,” “4G bonding,” etc.) that has left many people unaware of why engineers hit major workflow problems in the field.
Let’s take an initial look at where cellmux devices fit into the workflow today:
SOURCE AND TARGET (LIVE/FILE/WORKFLOW)
Cellmux technology is usually an option where either time or location determines that a satellite link (or local fiber) cannot be provided, and as such the typical cellmux use case is a fallback option usually limited to video workflows that require backhaul of video from ad-hoc locations. Typically this is the case for sports, news, or security applications. This might be live for linear insertion into a broadcast feed, or it could be file-based for post/pre-production insertion into an editorial workflow. The sources are usually SDI or similar, and the target workflow either expects the source video delivered into its digital IP-based workflow or will require that the video be converted back into SDI (or HD-SDI or similar) as it completes the backhaul to the studio.
Video compression in the cellmux devices is performed by either a DSP or FPGA chip, which is dedicated to the task and uses less power than the alternative, which is software or CPU-based encoding. Note that CPU encoding is highly configurable, while DSP, which is cheaper and uses less power, is relatively inflexible. While H.264 is dominant (largely because the codecs are now relatively mature), several devices have
introduced H.265 encoding (or will soon). H.265 is more CPU-intensive, but it is designed to produce more compression and lower bitrates at a quality equivalent to H.264. In bandwidth-constrained environments, such as backhaul, H.265 could introduce increased quality from deeper in the field. Devices might produce adaptive bitrate or variable bitrate encoding options.
Understand the networks you use. No two towers are the same, and the world of connectivity is complex and diverse. Try to understand and test the towers you will use at different times of the day.
Cellmuxes employ two dominant muxing/ transmission modes: RTMP (FLV), typically for the contribution to web workflows, and MPEG-TS for broadcast workflows. Both of these rely on TCP transmission provided by the channel-bonding link-aggregation processes.
CELLULAR AND OTHER BONDING
Each vendor has its own proprietary way of enabling its muxing. To get an idea of how this has evolved, let’s look at Link Aggregation Control Protocol (LACP), which is an established IETF standard and has been around (included in most Linux distribution) for many years. Each vendor’s proprietary link aggregation system presents different APIs and interfaces, but the underlying principles are much the same. In terms of reliability, this process is mission critical. Errors here can have widespread effects, so this piece of software must be written properly and is the crux of the cellmux’s unique capability.
Cellmux systems consist of two parts: the field kit (muxer) and the studio kit (demuxer). The demuxer comprises the counterpart to the muxer, reviving all the differently routed video packets, recombining them in software, and then delivering the combined video as a single coherent stream. This stream can be forwarded back onto the internet as a single stream, to a server for distribution to web pages, or as a secure targeted point-to-point unicast to a playout or studio where it can in turn be decoded back to an HD-SDI stream for use in a traditional SDI-based TV workflow.
Considerations for reliability include the physical location/security of the unit, the power supply, the hard drive spindle (mean time to failure), and of course the connectivity. Obviously the combined input and output bandwidth available to the demuxer must be at least double the expected maximum aggregated source bandwidth since the content needs to be aggregated in AND delivered onward. This typically means that the demuxer needs to be hosted in a data center with high speed internet access when delivering MPEG-TS or RTMP over IP. If the output required is SDI, however, then that high-capacity SDI “tail” from the demuxer to the eventual playout location could be prohibitively expensive, and it might be cheaper to locate the demuxer in the studio and bring in a suitable leased line. This linkage is also mission critical, and so choosing the demux location is a paramount part of the design.
Cellmux backhaul can combine VSAT with cellular. This keeps SLA as high as possible, while also offering lower cost when cheaper links are available.
DEMUX OUTPUT WORKFLOW
Once the RTMP, MPEG-TS, or SDI video is delivered and treated (if it needs re-encoding), the connection to the content delivery network has to be extremely reliable. Any issues with this link will be replicated to all viewers through the CDN.
Backhaul radio signal path
Those of you who know the cellmux space will have turned to this article expecting advice on antenna positioning. Naturally the physical and link layers (0, 1, and 2) are obvious targets for an article on cellmux reliability. While you could write a book on radio frequency optimization, for the purposes of this article these are some of the key things you should think about when planning your backhaul.
The first key task is to swallow your pride and to accept that even if you just spent a large amount of money on your cellmux, there is absolutely nothing you can do to guarantee the reliability of any radio signal, be it 4G, 3G, Wi-Fi, or even satellite. Satellite is much more predictable than the others since it is possible to guarantee available satellite capacity. Even so, satellite radio frequencies are affected by specific densities of rain and precipitation, and under these conditions even a satellite uplink with line-of-sight to the satellite can be rendered useless by the right density of fog; an effect called rain fade.
Assuming you are using a cellmux because no wired connection is available, try to get set up with Very Small Aperture Terminal (VSAT) satellite links, which will provide you guaranteed bandwidth (rain fade aside).
Your cellular services on the device should be treated as a fallback if wired or VSAT connectivity fails. Cellular networks will not sell roaming leased lines; you have to share all the cell towers’ backhaul with all other data on an equal footing.
However VSAT and wired connections will not always be options on location for all sorts of reasons. In fact there are places where a cellmux is the only device capable of providing any form of high quality backhaul, such as from boats, moving vehicles, or indoor locations where you can’t get a clear satellite signal.
Here are some thoughts and tips:
TOOLS TO KNOW YOUR DESTINATION
Site surveys and personal visits with the kit you plan to use are obviously the best option. If that is not possible, then Sensorly is a great service that crowdsources mobile coverage. Using this type of data can ensure you arrive with suitable SIM cards for the cellmux that reflect the coverage you are expecting at the location.
ReelSolver's Tim Siglin and BlueFrame's Chris Knowlton discuss the challenges of streaming via cellular bonding in congested cell-traffic areas, and strategies for boosting performance, in this panel discussion from Streaming Media West 2017.
Cellular bonding solutions have changed the game for the venerable animal welfare organization, which is able to go live from all locations in all kinds of conditions.
What to look for in cellular multiplexing devices -- perhaps the fastest growing streaming technology.