The “megabytes per second” metric that is the crux of what we purchase “our Internet” for, is rapidly becoming a meaningless number. The Internet carries a number of native protocols, TCP and UDP. One has guaranteed delivery, and the other doesn’t. We start from there, and neither was developed with much timed media in mind. Video and other isochronous media on the Internet was an afterthought, but we’re paying for it using the old context of bytes delivered per second. Instead, we need either a replacement measurement context, or a way to qualify delivery for comparison's sake.
Textual data shows up when you ask for it, especially in a browser. Boom, it’s kind of there after you ask for it. Text-based web pages, or those with small still pictures on them can load reasonably fast, even at dial-up speeds. Video doesn’t work that way. Video lives in something called a time domain, and must be delivered in such a way that it’s not interrupted -- otherwise we complain. Sheer megabytes/sec is meaningful only when you’re downloading a video for later viewing—and temporarily owning or renting a video is problematic for most content licensors. Realtime video needs a mostly uninterrupted path to the machine where it’ll be viewed. Video is isochronous data and you can interrupt it in only a few ways -- briefly at that -- or its display becomes objectionable.
But we don’t pay for that.
Retail consumer broadband usually delivers a specific amount of ostensible bandwidth expressed in megabytes per second, and we usually buy them in gradients. The figure is often quoted as so many megabytes per second down (as in download) and so many megabytes per second up as in upload. Some online organizations like DSLReports allow you to test the speed of your broadband connection to see how fast it is, then compare your number with the results from others having taken the same test, often searchable by zipcode. The number might be an optimized one; several iterations of the test you might try are recommended.
Sadly, the test results are moderately useless for video. Instead, there’s a trial-and-error experience for users that click on Hulu, NetFlix, and other online video resources. There are numerous influencing factors that make video dicey, even though it’s in high demand, and getting higher. Few broadband providers will tell you that all of them can be of high quality, or even guarantee them. Most of them can’t guarantee results, even though we desire them.
Sometimes an ISP or service provider has no control over the whims of the path between the service you desire and your ultimate playback machine. Other times, you may actually have an unoptimized connection and not even know it; this means that your specific connection is just substandard due to things like rust, or misconfiguration. These things happen.
The speed test you took for your broadband connection might have been one that sent data across the country. In video realtime downloads, the role of content delivery networks (CDNs), or networks within networks, plays a key role. CDNs are geographically dispersed server farms that perform one mission in life, and that’s to deliver content, usually video, to consumers. They’re geographically dispersed to make the number of router hops to you as few as possible but to as many others as rationally possible.
This is because they can handle only so many concurrent streams of video themselves until their pipes are saturated and the streams will cause delays. CDNs must also fight biorhythms, that is, the daily demand cycles that occur through the day for downloads that might contend with video delivery quality. They also must service the equivalent of the ADHD remote control holder: the happy video clicker.
Video consists of frames that have to be timed together. Internet data delivery wasn’t designed that way. There are some protocols called Quality of Service (QoS) protocols that try to clear pathways between video source and destination. These work up to the point where network congestion screws up timing, rendering “buffering errors” that translate to poorly perceived video quality at its destination.
But we don’t say: Hulu Approved. Netflix Guaranteed! Part of the reason is that Comcast’s xfinity network competes with both of those, Apple, Amazon, and anyone else that wants to deliver video to you. AT&T, a major DSL supplier is in the same boat, as are other ISPs and service providers. They want to sell guarantees and satisfaction for their own wares -- their own video on demand services. So the impetus to provide competitive metrics on video delivery over Internet wires has few industrial motivators. It’ll be the third parties, not the first and second parties in the service ecosystem that press for this. Oh, and we, the civilians and tech community that press for usable comparison metrics for video infrastructure delivery.
This story, "Broadband Bytes Have Lost Their Bite" was originally published by ITworld.