A Buyer's Guide to Cellular Multiplexers 2014
As part of a project I began for Streaming Media 4 years ago, I’m now currently tracking 16 vendors of cellular multiplexers (as well as a handful of channel-bonding and link aggregation technology providers). The number of vendors is up by four from last year and up by 11 since 2010!
That is quite significant growth for any niche in this sector, apart, perhaps, from the online video platform rash that had an outbreak about 5 years ago and promptly disappeared.
Especially notable is that each of the vendors that has entered the market has at least one strong differentiator from the others. In my experience, vendors in these streaming media subgenres can be very catty about the others’ products, but the cellular video multiplexer (cellmux for short) community is generally very mature when you talk about the pros and cons of one particular product over the others in the space. This indicates to me that the vendors are finding that the competition is only helping to grow the market and that the market growth is outpacing the growing number of products and vendors. This, of course, is a healthy thing for everyone, particularly the buyer.
Let us take a quick look at the basic product types, with a few comments about the features that you should be thinking about when making your buying decision. First, let’s make sure we understand the products.
Cellmux is a combination of a video encoder and a channel-bonding link aggregator. “What the hell is that?” I hear you cry.
A link aggregator is a device that can connect multiple Layer 2 network links (for example, ADSL, Ethernet, Wi-Fi, and fiber together and aggregate the bandwidth in such a way that any requests can be load balanced across all of these available connections. For web traffic or burst email and so on, a link aggregator can effectively allow a group of people using many applications to always find an uncongested route over which they can reliably send and receive their chunk of data.
Critically, a link aggregator only needs the one device at the customer premises, and nothing need operate remotely in the network. Each request is load-balanced and sent via the least congested route. For burst data, this gives the user greater network availability, increasing the ratio of bandwidth per user. However, if the highest bandwidth connection is (for example) 1Mbps, then the user who “gets time” on that connection may stream video up to 1Mbps, but even if there are 10 more 1Mbps connections, the fastest video they will be able to stream will still only be 1Mbps.
In order for a consumer to exceed this 1Mbps limit, a secondary stage needs to be added. A remote device located in a hosting center that has (for example) 100Mbps connectivity could work in conjunction with the consumer’s link aggregator to work out how to split chunks of (for example) a 4Mbps stream over several of the link-aggregators’ 1Mbps pipes. It would then coordinate reconstituting these separated chunks of video back into a coherent, single, higher bitrate stream. This process is a form of “multiplexing,” and in this model, where the Layer 2 links are often of different types, the common term is “channel bonding.”
So a channel-bonding link aggregator has two parts -- one in the consumer’s location and one in a well-connected hosting location.
Channel-bonding systems work well and have been around for many years. In 2003, I recall using ISDN channel bonding to combine six 128Kbps links into a single 768Kbps “pipe” for streaming 512Kbps quality video.
However, when you move off fixed QoS connections onto Wi-Fi and mobile connections, channel bonding is prone to variation. If you set up your video encoder to produce a fixed 512Kbps, it is highly likely that, at times, only a few channels will be available -- so the aggregate bandwidth will be insufficient to continue to stream your 512Kbps source stream. Now, while we are all familiar with adaptive bitrate streaming, typically this is something implemented between the server and the end users’ media players to help to deal with these volatile network conditions of the end users. As most readers who produce webcasts will know, the connection between the origin encoder and the origin server (from where the distribution networks source their signal) is imperative. If your source from the encoder to the origin server is bad, then every user experience is bad (also known as a garbage in, garbage out, or GIGO, process). Accordingly, it is very important that if the mobile encoder is passing through areas of variable signal coverage that the video between the encoder and the server must also be adaptive. To ensure this, there must be some computational communication between the network layer channel bonding link aggregation and the video encoding process.
Broadcasters will be able to stream live content to multiple devices using a cloud-based workflow.
Forget the sexy form factor and look deep inside the technology. Here are the three questions to ask before buying a cellmux.