So yesterday Dan Caruso took issue with my post on measuring online video growth , but then we had to wait a whole day before he said why. Yes Mr. Powell, I’d like you to meet Mr. Hanger, Cliff Hanger. Now given Dan’s earlier positions that bandwidth growth is in fact very healthy, I thought at first he was going to dispute the word ‘if’, but it turns out he was looking squarely at the word ‘triple’. Here is what I said:
Folks seeking to measure the growth of traffic from online video have to look past the videos viewed and take into account the changing parameters of those videos. The explosion, if it happens, will come from more videos being viewed at higher average bitrates and for longer times – a triple effect
What Dan presumeably means is that if people watch twice as many videos at twice the bitrate for twice as long, the effect is not ‘triple’ it is 8x and the growth is exponential rather than linear. And exponential growth would surely lead to an explosion! And each effect can be increasing at its own rate, e.g. HDTV is several orders of magnitude above a YouTube video in bitrate if people start streaming it in quantity the growth of average bitrate is going to be much much larger. I should have known he’d look at it that way.
Now, I obviously wasn’t using the word ‘triple’ as a multiplier, only to count the effects we ought to look at, but I will grant that it was ambiguous. But hey, IT WAS A BLOG POST, some ambiguity is a given, I’m an engineer not a poet.
FWIW, I don’t think all the effects can be multiplied either, they are interdependent. For instance it doesn’t make sense to expect both the number and duration of videos to go up independently, if people watch a 2 hour full screen video, that’s 2 hours they weren’t watching 30 YouTube funny puppy clips or something. Longer duration leads to fewer actual videos watched. But I was ignoring such interactions and just counting the effects. I was also ignoring the *size* of each effect, I was just saying the other effects existed and should be considered!
So Dan, rather than the word ‘triple’, what word or phrase would you use?
If you haven't already, please take our Reader Survey! Just 3 questions to help us better understand who is reading Telecom Ramblings so we can serve you better!
Categories: Content Distribution · Internet Backbones
Join Our LinkedIn Group
Don’t forget MPEG 2 vs. MPEG 4.
And don’t forget that a little traffic engineering (which carriers have been doing since they only sent dots and dashes) using multicast and caching takes it down a lot.
Can anyone here state with confidence and precision the manner in which the video traffic stats being discussed in this thread are actually obtained? If so, please point to the rules set, or briefly describe them.
Surely various and often widely different sets of criteria are used across different venues of interest to determine the amount of traffic traversing not only the core, but edge networks horizontally and cached, as well, and some of the latter are never included in some core-based analyses. I’m asking this because of an earlier observation I made, which I posted to my forum at the URL below, which now causes me to wonder which criteria were used in the topic at hand.
http://tinyurl.com/696sc5
[for brevity, you can skip to the paragraph contain a a bold-typed sentence]
It doesn’t matter anyway. Elliot Soft-Spitzer says price declines will nullify bandwidth growth…so all if for naught…us bandwidth bigots are just running on an increasingly fast treadmill
I think we’ve established now that growth *is* occurring, bandwidth growth is outpacing pricing declines now. But it is very early yet in the game. I think the next great challenge though is finding a way to make costs go down with those prices. If a 40G circuit is not cheaper than 4x10G and the same happens with 100G, then that treadmill is going to get painful real quick.
There has never been a question as to whether video growth is growing or not. What is paramount here is the ability to determine where it is occurring, how it is measured, by whom, and the criteria being adhered to.
Conceivably, video growth, as a function of end user viewing habits, can be growing by several orders of magnitude, but if a paradigm shift were to occur in respect to ‘where’ content is being stored, cached or otherwise stored and forwarded, such traffic could easily (and I suspect currently does) elude capture by most of today’s measurement practices.
This goes back to Andrew Odlyzko’s and my earlier comments concerning traffic that is either sequestered within walled gardens, or exists in peer-to-peer form at the edges prior to or lieu of gatewaying onto the larger, public Internet, where ISP-Edge routers do the bean counting.