Slashdot Mirror
Netflix Compares ISP Streaming Performance
boustrophedon writes "The Netflix blog compared streaming performance among 20 top ISPs for the past three months. A Netflix HD stream can provide up to 4800 kbps, but the fastest American ISP, Charter, could sustain only 2667 kbps on average. Most Canadian ISPs beat that, with champ Rogers providing an average of 3020 kbps. Clearwire, Frontier, and CenturyTel were in the doghouse with under 1600 kbps."
I've got Verizon FiOS, and though I know it's not that common, but I can get steady 3.7 MB/s streams.
I'm not going to suggest that you are incorrect, but I am going to suggest that your single piece anecdotal evidence is not nearly enough to discredit the report Netflix put together.
Is there a value to looking at "averages" over large areas?
I'm in Howard County, MD. On Comcast, I get about 8Mbps sustained after an initial 20Mbps. This is typical in my area. My neighbors using Verizon FiOS will typically see even higher throughput. However, this graph, presumably containing large parts of the US, has Verizon as slower than Comcast, and both are much lower than what I see. A person trying to make a decision on ISP service in my area would be misled.
Very nice. Rather impressive to pre-empt the ISP's.
"well, your competitor is able to provide better speeds to more customers, why are you whining? Oh? AND You charge more for lower service? Interesting. Well, lets let your customers decide for themselves with more facts who they want"
It'd make sense at this point for an ISP with a bit of sense to make a nice deal with Netflix to improve things here, then everyone wins.
Waiting for an amusing sig.
Yeah, my cable modem is stupid fast....only problem is, running full tilt I can go through my monthly bandwidth cap in eleven and a half hours.
Fortunately, for the moment, the overage cap is $50 so if you download a bunch some month you just say, "Woohoo, unlimited bandwidth." For example, in January I downloaded 750MB which put me 625MB over my cap and would have cost an extra $780. Ridiculous no? And now the CRTC (equivalent of FTC) has ruled that the major ISPs are allowed to pass usage based billing fees onto third party providers which means there will be no more unlimited plans and the billing cap will likely go away.
Basically, Rogers and Bell want you to watch their channels, not use Netflix, AppleTV, etc. And the wretched hive of scum and villany known as the CRTC is letting them do it.
Not much point in fast internet if you can't use it.
In Soviet Russia, hot grits put YOU down THEIR pants.
There's Verizon FiOS and Verizon DSL. Is the measurement for FiOS, DSL, or both?
.
Prisencolinensinainciusol. Ol Rait!
The article uses Mbps (megabits per second) and not MB/s (megabytes per second).
3.7MB/s is a LOT of bandwidth, it's 30Mbps. Not even FULL HD video uses that much bandwidth.
You probably meant 3.7 Mbps.
...but how in the hell is anyone supposed to pick the colors out of those graphs, at least three of them are the same shade of sky blue.
I'd like to see this redone as the graph is certainly compelling, just a little bit more readable.
If the world isn't beating a path to your door you're doing something wrong.
Fantastic. The worst ISP in Canada is still faster than the best ISP in the US.
Also, while interesting, this is basically useless to the average US consumer. It's not like you get a chance to choose between those 16 US ISPs. In the US, you're lucky if you get to choose between 2 of them.
Fiber optical very high speed equipment (used behind the xDSL/Cable copper network and behind the wireless towers) has never been as cheap as its now.
Long range gigabit ethernet stuff is dirt cheap.
10 Gigabit ethernet, which allows 2 thousand 4.8Mbps streams is already very affordable for carriers.
A pair of fiber strands can carry 16 10 Gigabit links easily, that's enough for 32000 top speed streams.
Long distance fiber optical cables typically have at least 36 strands. Some reach as high as 144 strands. Do the math and you'll see fiber capacity is almost never the issue.
But then there's a very interesting FACT. If you exclude p2p and video streaming, 10Gbps link can provide bandwidth for one million users. That's right, typical users that don't run P2P or use video streaming/download services require just 10kbps bandwidth.
That's the conundrum. Heavy users (p2p and video streamers) are responsible for such a disproportional share of bandwidth consumption that most providers just don't care about the quality of service they deliver for such users.
They might actually prefer you go elsewhere given the tight margins.
You should be glad if you can pay US$ 10 more to have higher quality service. At least that way you ISP can't complain that your piggybacking on the average light user. Honestly asking for a discount if you use little bandwidth isn't very reasonable either. A very large portion of the ISPs cost is last mile stuff that doesn't change if you're using zero bandwidth or maxed out in the xDSL case. Copper / coaxial cable stuff requires most maintenance, as they are most subject to ice storms, lighting, traffic accidents, ... Fiber can have redundancy.
ADSL is the best option for video, as long as you're close enough to your ADSL provider DSLAM that you can get fast enough speed. Cable is only better if you're far from the nearest ADSL DSLAM. DSLAM is the counterpart to the ADSL modem. Your ISP should be able to estimate your max link speed with your address and tell you when you're too far for your selected speed.
Just trying to demystify some facts. Most large ISPs / carriers won't discuss this stuff openly. They prefer you don't know the real facts. I'm not trying to judge what is right or wrong from either side.
The different connections need to be split.
For example, Verizon needs to have:
Verizon DSL 768kbps - 1Mbps
Verizon DSL 1.5Mbps - 3Mbps
Verizon DSL 4Mbps - 7Mbps
Verizon DSL 10Mbps - 15Mbps
Verizon FIOS 15Mbps
Verizon FIOS 25Mbps
Verizon FIOS 50Mbps
Obviously a low end DSL connection is not going to be the same as those who can order the 10-15Mbps DSL connection. And it is likely that the DSL 10-15Mbps connection is going to be different from the FIOS 15Mbps.
To group all of those connections into one Verizon line is completely misleading. And if they didn't take measurements from all of those connections, then then that makes the results even more suspect as the graph doesn't specify what type of connection they chose to test with.
Full HD streams using H264 run around 10Mbps for poorly compressed video (small publishers without a video engineer doing the work).
Look at wikipedia:
http://en.wikipedia.org/wiki/High-definition_video
It says that the heaviest video streams from Netflix use 5Mbps.
I just played a few of the 1080 Full HD streams I have. The highest bandwidth they use is 10Mbps.
30Mbps for a video stream is completely crazy, even if your ISP service is that fast, you won't be able to sustain that speed, there will be serious drops.
Only Blue Ray uses that kind of bandwidth. Not even DVD.
Maybe he's talking about video content accessed directly from Verizon. Then it's not internet streaming video, it's VOD content coming directly from your ISP.
It's not ranking YOU. It's ranking them as a company, which includes all tiers of their service.
Better yet, the FCC should be running the test.
"The Consumer Broadband Test, currently in beta, is the FCC’s first attempt at providing consumers real-time information about the quality of their broadband connections."
http://www.broadband.gov/qualitytest/about/
The larger the ISP, the more they’re penalized by the more rural regions which are limited to DS3 45 Mbps circuits feeding a whole town.
The AT&T numbers will suffer somewhat due to Uverse (VDSL) and regular DSL being sold at varying speeds. While I like the idea of 18-24Mbps, I can't reasonably afford it at this time.
-l
Help cure AIDS, cancer, and more. Donate your unused computer time to worldcommunitygrid.org. Join Team Slashdot!
Exactly what I was thinking for Verizon and maybe some others... A lot of Verizon's DSL connections are either 3.0 or 1.5 capped which would destroy their throughput...
that said probably 80+% of Verizon's service IS DSL not Fios, so its actually probably lifting their numbers higher then they should be, not the other way around. Last I heard Verizon had stopped rolling out Fios (this may just be internet rumor mill crap)
The biggest problem with this isn't the ISPs, but the customers. I was working for a nationwide cable ISP for about 3 years, and we offered speeds of 1.5Mbps, 3Mbps, 12Mbps, and 25Mbps. Which do you think the majority of people chose? The 1.5Mbps for $15-$20/Month. Most people didn't want to spend the extra $15/month for the 12Mbps connection, and then complained about the problems when they tried watching 3 streams from Netflix and online games at the same time. There's no way they would up their speed though, it's the ISP's fault that the speeds were so slow. There's just no winning when everyone goes for the cheapest product available. It's the same reason Walmart is doing so well... Most Americans go for low price over quality.
only problem is, running full tilt I can go through my monthly bandwidth cap in eleven and a half hours.
We badly need a "truth in advertising" law that would make it illegal to label a "100Mbps connection" with a 5GB monthly cap as anything above the 16331bps it really is (yes, less than 16kbps, this is not an error). Providing a bigger burst is ok but only if that's clearly marked as such.
Toss in something about the scam that lets ISPs call 100Mbps down/128kbps up by the bigger number. If you want to use just one number, you'd need to print the lower one. Anything else is deceiving the customer.
The creatures outside looked from Alt-Right to Antifa; but already it was impossible to say which was which.
Netflix streams at rates that effectively consume the entire subscriber subscription rate for hours on end. It is obvious from the postings which are being made that most who are posting are end users with no network design knowledge. So, here goes.
Netflix streams at 1.0-6.0 Mbps at a fairly constant clip the entire time a movie is playing. There are exceptions to this behavior. But, for the most part, that is the picture the network engineering folks see. Now, take a rural telco of 5000 access lines that offers DSL to 100% of its customer base. Today 2000 of the 5000 lines are subscribed and running ADSL2+FAST and 90%+ can attain speeds of at least 3 Mbps. Most can attain much higher speeds.
Of the 2000 DSL subscribers, lets assume that only 1% are Netflix customers (arguably a low number). If 1% are Netflix customers, then twenty customers will be consuming 60Mbps of bandwidth at 100% capacity constantly. If 10% are Netflix customers, then 200 customers will be consuming 600Mbps of bandwidth at 100% capacity constantly. This rate is twice the rate of the purchased bandwidth capacity for the entire ISP. And this assumes only 3 Mbps service.
Now lets assume that those 200 customers are paying $29.95 for their 3 Mbps service. The total revenue for those 200 customers would be $5,990.00 per month. That amount is almost $3000.00 per month less than the cost of the upstream port capacity without any transport fees. Oh and the bandwidth being purchased for the entire ISP is only 300Mbps redundant capacity. BTW - There is no regulatory recovery for Internet bandwidth. So, please don't play the telco monopoly card.
So, there are a bunch of end users who think you are getting the short end of the stick from the ISPs. Think again. Just because Netflix wants a free ride to the end customer and the end customer wants to use 100% of their purchased link on the cheap doesn't make the ISPs the villians. With this kind of network utilization, the ISPs are no longer able to make any money on their services. As well, they are now having to rebuild their networks to make sure they can support over the top video services like Netflix. And, Netflix is rating the ISPs???!!! That is like the fox guarding the chicken coop. Why would anyone want to hear from Netflix how well ISPs perform for their service when Netflix isn't paying for the network or bandwidth.
A smile and a hug never hurt anyone
label a "100Mbps connection" with a 5GB monthly cap as anything above the 16331bps it really is
I completely agree.
I'm getting more and more pissed off as greater numbers of people are beginning to understand what monthly transfer caps are, and then proceed to voluntarily or forcedly believe the outright fucking lie that these assholes are perpetuating (and that most folks here also believe!):
[bullshit]
Enforcing a limit on data transfer over a given period of time is a very direct and extremely effective method of completely alleviating the problems that can be caused by a small number of users consuming most or all of the bandwidth on any given, shared network segment.
[/bullshit]
The statement that you just read, enclosed in [bullshit] tags, is 100% bullshit.
It's a problem with bandwidth, not a problem with transfer. Don't ever believe the utter lie that these two concepts are inherently and directly correlated. While they can be correlated, they do not have to be, and, in true Slashdot spirit, they are most certainly not causal.
Lastly, if you don't understand WHY what I say is true, think of it this way:
Take a look at the gigabit switch sitting on your desk (or pretend you have one). You've used it very lightly. You just browse the web through it. Maybe some games. No youtube, no torrents, no downloading. You've owned it for a while now, and you've transferred about 10 gigabytes of data through it in that whole time.
I own the exact same model, and I'm coming over your house later and swapping out switches with you, but the difference between your switch and mine is that I pushed 5 terabyes of data through mine every single day I've owned it.
Given that neither switch is defective, when I switch hardware with you, will you notice the difference?
The worst part about this whole thing is that bandwidth is a fundamental commodity and property of multi-segment interconnected networks in general (read: the Internet). It's so fundamental that, rather than paying specifically for the connection speed of a physical link into someone's network, ISPs pay specifically for bandwidth usage based on a well accepted model commonly called 95th percentile billing because of how fairly and accurately it reflects a given link's impact on the network. Overall transfer over a given period, while it may be calculated, is irrelevant because the amount of data pushed through a link simply doesn't fucking matter. Data transfer at all levels of a network is a function of bandwidth, not the other way around. Were you to graph it out as a function, as time approaches infinity, transfer does as well.
On behalf of the ISPs though, this misconception and billing model is absolutely genius. If I literally possessed a LIMITLESS source of product (data transfer) and, irrespective of size, somehow managed to convince you that it was reasonable for me to charge you for a finite, expiring quantity of it, I'd laugh all the way to the fucking bank every time you came back for more.
Boot Windows, Linux, and ESX over the network for free.
Is the real offense here by Netflix, as evidenced by comments there on the site, is that they've exposed the dirty little secret of many ISPs--that they oversell capacity?
Dude, where's my packet?
Better graph: http://waxy.org/2011/01/colorblind_leading_the_blind/
Netflix is paying for the bandwidth on their end, so it's not like they're getting anything for free. The issue is that all of the sudden people want to stream massive amounts of information and it's breaking all the assumptions that went into oversubscription ratios.
The answer is for ISPs to be honest about what they provide. If there's a monthly cap, say so.
That said, it would be good for netflix users to be able to pick a speed and have it stay there rather than sucking up all available bandwidth.