Will Netflix Destroy the Internet? 577
nicholasjay writes "Netflix is swallowing America's bandwidth and it probably won't be long before it comes for the rest of the world. That's one of the headlines from Sandvine's Fall 2010 Global Internet Phenomena Report, an exhaustive look at what people around the world are doing with their Internet lines. According to Sandvine, Netflix accounts for 20 percent of downstream Internet traffic during peak home Internet usage hours in North America. That's an amazing share — it beats that of YouTube, iTunes, Hulu, and, perhaps most tellingly, the peer-to-peer file-sharing protocol BitTorrent."
OK, and? (Score:5, Informative)
First, they've known this was coming for ages. P2P have been around well over a decade and everybody knew people were downloading movies and TV shows and watching them on their computer. It's just hitting bigtime mainstream now and Netflix was the first commercial entity which did it right.
Second, the 'Will Porn/Youtube/Torrents/P2P/Netflix/etc Destroy The Internet?" articles have been around for ages. The providers adapt, the technology adapts.
How does never work for you (Score:3, Informative)
Most ISP's in the US already have a (high) data cap. Whatever you do under that, they will not care. If there were (or are) any ISP's with "unlimited" bandwidth then they will have to change policy also to have some kind of data cap, because they do not get "unlimited" bandwidth from the people they purchase internet connectivity from.
Already a non-starter in Canada (Score:5, Informative)
The average person has a 60gb cap in Canada. People have quickly found out that they can blow through 1/2 to 3/4's of their monthly cap in a weekend. I'm sure it'll be more interesting as winter rolls around, we like snow, hockey, and all that. But curling up to watch a movie or 4 when it's -40C and snowing out is much better fun. Especially if there's a 30% chance you're going to spend 3hrs shoveling.
But sandvine is a blight on the internet. You can happily hear about all the horror stories(look on dslreports.com) that they've inflicted on Canadians, as ISP's use their equipment to throttle just about everything. Bell enjoys using them after the last mile, before switching to outside networks, even when you're on another ISP. So regardless of what happens, you're still being throttled by bell. Rogers like using it to throttle everywhere, that they think the consumption might be too high, or where growth is outpacing their delayed upgrades.
Re:So, how long before... (Score:1, Informative)
Umm, the internet was designed to be a redundant communications network to keep military bases connected. It was to address the concern that a nuclear war would knock out communication to underground bunkers at the Pentagon, Cheyenne Mountain, and SAC HQ.
It was never envisioned to be a commercial or entertainment vehicle.
Re:So, how long before... (Score:3, Informative)
Comment on the statistics (Score:5, Informative)
Farhad, Allow me to make one clarification on the Sandvine report cited. While the growth of Netflix has certainly been dramatic, it does not (yet) account for 90% of Internet traffic on any of the networks included in our study. Rather, As you noted correctly, we did see Netflix accounting for approximately 20% of downstream traffic in North America.
The confusion on the 90% stat probably resulted from a misreading of one of the graphs featured in our “Spotlight On: Netflix” on page 15 of our Fall Global Internet Phenomena report. The graph was accompanied with the caption “An average day for Netflix on this network, peaking at 9:30pm” This particular graph (taken from a single network in Canada) shows Netflix traffic throughout the day as a relative percentage of the peak amount of Netflix traffic. In this case, the peak was reached at 9:30pm, so the curve at that point has a value of 100%. The rest of the curve shows how Netflix traffic varies: so we see that at midnight the level of Netflix is approximately 42% of what it was at 9:30pm. In hindsight, I think we probably could have explained this better in our report.
Our Network Analytics product produces these “Time of Day” graphs so that network operators can understand how subscriber usage of various applications, services, or categories of application vary throughout a typical day. Thanks again for the interesting article.
Sincerely, Tom Donnelly, EVP Marketing, Sandvine
Re:Streaming Netflix was disappointing (Score:3, Informative)
As TFA implies, selection has improved greatly over the past year or two.
Re:So, how long before... (Score:4, Informative)
that was supposed to be only for the testing group
problem is it never ended
http://www.networkworld.com/community/blog/why-ipv6-vint-cerf-keeps-blaming-himself [networkworld.com]
some wanted a 128bit others a 32bit..
Re:Bandwidth? (Score:3, Informative)
Yeah. If the regulators had had their way we'd all have ISDN by now.
Re:How does never work for you (Score:5, Informative)
A classic study would be Canada. When Netflix came to Canada or announced plans to do so, Rogers (cable) immediately LOWERED their measly caps from 60GB to 30GB-ish. Bell (DSL) heavily lobbied the CRTC so DSL connections can be billed by the byte, so that ISPs using Bell's lines are at a huge disadvantage. Shaw (cable) already announced plans to charge overage charges at $2/GB (for "lite" and "high speed" users) or $1/GB (for the faster plans - warp/nitro). You can pay extra for more - $5 for 10GB and the like. Right now it's a trial, but they're planning on rolling it out.
SO yeah, Netflix's potential for clogging the Internet won't happen. Ditto the "bandwidth crunch".
Heck, the FCC may make stupid rules, but in Canada without those rules, things are a lot worse. I can't have digital cable without buying and paying monthly fees on the cable provider's box (which only works with that provider - they won't (and don't have to) allow activation of 3rd party boxes). No Firewire video at all. No cablecard (crap, but at least I could use my TiVo). No unencrypted digital cable (if I want high-def for free, I have to stick an antenna up - no rules saying all those channels must be unencrypted QAM), etc.
Re:The answer is... (Score:3, Informative)
Re:No, it's just static content. (Score:5, Informative)
Re:The answer is... (Score:3, Informative)
FYI, Netflix works just fine from Firefox running under Windows (at least it does on my home XP box). And no, aside from inside of a virtual host there is no way to watch Netflix instant streaming movies with Linux.
Re:So, how long before... (Score:2, Informative)
It was designed to connect research institutions (universities) to each other and to military research institutions (DARPA). It was not designed to be resistant to attack, much less nuclear attack. The ability to reroute is, and was, necessary because physical equipment and lines are prone to intermittent failure.
You, and whoever modded you Informative, are ignorant idiots.
Re:The answer is... (Score:3, Informative)
Seriously, is there a way to stream Netflix under Linux yet (aside from in a VM?)
Why bother? It's only 1% of the market, if that.
Watch it in a VM, or watch it on your Wii, or your Xbox, or your PS3, or your bluray player, or any of the scores of other devices that stream Netflix.
Yes, Netflix works in all browsers, it just doesn't work on Linux. And frankly, most people don't care.
Re:So, how long before... (Score:5, Informative)
Here is the problem. They want to MAXIMIZE profit from the bandwidth. Not get a good profit or healthy profit, but MAXIMIZE it in any way possible. Comcast does it by intentionally not upgrading their downstream paths. Even 10 years ago Comcast was capable of 10BaseT speeds Up and Down over cable modems to the headends for ALL the people in the area that headend serves. The problem is that headend is connected via fiber to a larger headend. That larger headend has another 5-10 connect to it, and a Single OC3 feeds 5+ cities if you are lucky for it to have an OC3. The area I worked in was selling 5Mbit service and I knew that the backend was nothing more than 2 bonded t3's that way too little bandwidth for the number of subs on that POP.
ISP's are screwing the pooch in increasing their backbone connection speeds. Until they get a LOT of complaints, they will continue to major oversell the available bandwidth. it's now well past the 100 to 1 ratio at most.
Re:I think he means things like cache engines (Score:5, Informative)
Netflix uses Akamai.
Re:How does never work for you (Score:1, Informative)
I work for Shaw so this is being posted AC for obvious reasons.
I wish to address this so as to dispel any misconceptions here.
*Shaw has always had bandwidth caps, they have always shown those on their websites.
*Shaw never marketed as unlimited internet. In the beginning it was always "always on connection".
*Before these bandwidth caps if you went over and the node was saturated you were turned off.
*Shaw has always addressed saturation issues quickly and promptly if you are in a major city (if you are in the middle of no where like Souix Lookout Ontario it may take half a year).
*When Rogers reduced their bandwidth caps, shaw increased theirs. Regular internet went from a 60GB cap to 75GB, Extreme went from 100 to 125.
*Yes while there will be overage charges customers can pre-purchase allocated bandwidth in "packs" which are MUCH cheaper.
So since Shaw's only actual product is communications. They have to bill based on this, big surprise. There is plenty I don't like about Shaw, but this situation is not one of them. Shaw is not perfect but apart from perhaps TekSavvy they are the least evil communications company in Canada.
Re:So, how long before... (Score:5, Informative)
A lot of people really don't understand that. They probably have cheap hosting accounts, and have never needed to deal with actual circuits.
I've worked at places with multiple GigE circuits. Besides the base cost in the datacenter (floor space, power, port charges, etc), we had negotiated contracts on bandwidth. 95th percentile is that bastard of a number that we deal with all the time. For those that don't know, it goes something like this. The uplink provider monitors our ports once every 5 minutes. At the end of the month, they take all the samples, sort them by utilization, and knock off the top 5%. Whatever that next number is, is what we pay. There are dedicated rates too. If you don't use the line, that doesn't matter, you're paying at lest a minimum fixed amount, which could be something like 20% of the line capacity. So an idle datacenter with nothing in it, but it has a GigE circuit could cost as if we were using 200Mb/s at 95th percentile.
For our bills, it was easily over $100,000/mo. That's a conservative number, but I haven't been there in a while, and don't remember how high it really went. Do you want fancier services, like multiple circuits into your space, BGP routing, etc? Oh, the price goes way up. When you get big enough, and want to get your data to the customers faster, you start doing private peerings, and putting out edge nodes (servers closer to the clients, like Akamai provides), or even putting dedicated servers in on the end user networks. They don't like paying huge bandwidth and peering bills, when they can deploy $100k worth of equipment two hops from the customer.
If NetFlix is sucking up so much bandwidth, someone's making a fortune on it already. So it accounts for 20%, big deal, that doesn't indicate the total utilization of the available, or even where it was measured. I played this game once. I took the total bandwidth my company used during peak hours, and compared it to the Mae East bandwidth graphs (when they were public). Our bandwidth used 15% of what Mae East passed. And guess what. It didn't destroy the world. We weren't even responsible for 15% of what passed through Mae East, because various peerings meant our traffic went in all kinds of different directions.
By that standard, NetFlix could use 200% of what passes through Mae East (plural now), and even that wouldn't mean anything other than bragging rights. Sure, it's a lot of bandwidth, but it doesn't indicate saturation of available resources, nor the end of anything at all.