What 1.6 zettabytes might look like:
http://drivebytes.anodal.org/#1.6_ZB
Networking megafirm Cisco has released a numbingly comprehensive report forecasting IP traffic up to and including 2018, which comes to the unsurprising conclusion that the market from which Cisco earns its bread and butter will continue to balloon for the foreseeable future. The Cisco Visual Networking Index (VNI) contains a …
Actually I'm confused by the stat of 5 million years per month.
That means there will be a total of 60 million years per year in 2018, or 60 million unique streams at any one time.
This doesn't sound very much, considering the number of people accessing video this way - I'd have thought 10 times that would be perfectly believable.
Reg, is there something amiss here?
That is certainly a horrible way to portray how much data is consumed per unit time. We can convert it to something more comprehensible though.
First, I'd take that 5M years of video per month and get the number of concurrent streams that need to be consumed full time to fill that requirement. i.e. 60M months of video are consumed every month. However, that 60M is still the total amount of video consumed. It's still not a meaningful unit to most people.
The easiest way to consume 60M months of video per month would be to simply have 60 million concurrent streams running 24x7. Thus, 60 million people could be streaming TV all month to get that amount of video. However, IIRC, the average American watches about 2 hours of TV per day. Thus, only spending about 1/12 of the month streaming videos. So, that initial 5MY/mo figure is enough to support about 720M people using streaming video for all their TV needs.
...but we're not going to guarantee that they did not actually mean zebibytes.
I thought I had a handle on this stuff, and then I was foolish enough to search for zebibytes. Next thing, you'll be telling me that someone has worked out a prefix for units greater than yotta/yobi. I am going back to an earlier, more innocent method for counting: one, two, lots.
"It would take an individual over 5 million years to watch the amount of video that will cross global IP networks each month in 2018."
Good. This will keep the NSA from spying on so many people; or, alternatively, provide 60 million new government jobs at NSA to watch all that video.
As it'll still be exponential and predictable growth. As long as Cisco and other equipment manufacturers do their homework, all ISPs need to do is to steadily upgrade their network... which they should to do anyhow. It's what people pay them for.
Video on the other hand is even rather undemanding. It's completely delay tolerant and for video on demand you can even download the file and play it while it's downloading. You can unplug the cable for seconds before the user will notice anything. Compare that to VoIP where every single lost or delayed packet can cause degradation.
You forget LIVE video streams, which are NOT delay tolerant since the packets aren't buffered. Think the World Cup right now. Imagine millions of people using their browsers to watch a game live over the Internet. This is the kind of event where broadcasting makes sense, but the Internet isn't built that way. The end result: millions of live streams transmitted at several hundred kbit/sec each. And each game lasts a while, so expect this hammering to continue for hours. See the problem?