Fast.com now measures latency and upload speed
media.netflix.com
media.netflix.com
In fairness, this would be true of any popular speed test, and I'm not sure of an obvious solution. Any kind of tunneling or onion routing would render the test pointless anyway, so I'm not sure of the extent to which you can ever really trust a speed test of this nature. The only really useful data for this kind of analysis is likely to come from a router with sufficiently sophisticated software that it can report on transfer rates over time (which many do), and even then, that doesn't account for scenarios where the remote connection is the bottleneck.
It was introduced after reports of ISPs doing just that, and it was Netflix’ way of hitting back.
If not then they could very well be fudging the numbers for Fast.com and not Netflix proper.
It may not be common or even legal in the US (I really don't know), but in the UK traffic shaping is commonplace, even in instances where the package has been described as "Unlimited".
...ntflxvideo.net
...ntflxvideo.net
...ntflxvideo.net
...ntflxvideo.net
You can see various request to *.nflxvideo.net
I have no idea if this lawsuit is still ongoing, because Free seemed to have had a first peering with Netflix in April, and has been going up in the chart:
Put another way, it is my impression that consumers are already paying for bandwith. Getting Netflix to pay for faster routes would mean ISPs are double-dipping.
If they want "premium" routes for their MASSIVE traffic, they will have to pay to the ISPs. I think that's reasonable. Giving them premium routes for free would not be neutral, would it? ;)
That is fair.
They shouldn't pay for our downlinks, that we've already paid for.
That would be double dipping or something by our isps.
ISPs should provide connection to the internet for their customers in exchange for a monthly bill. How they do that isn't the customer's concern and it shouldn't be Netflix's concern. The only ISP Netflix should be paying is their own link to the internet (AWS last I heard)
Side note, Netflix isn't using any traffic. They're not sending me a UHD video stream unsolicited. I'm using the traffic.
Milk companies put their lorries in the highway to deliver their products. You're the one buying milk. Are you the one who's using the highways?
If Netflix wants a premium highway let them pay for it. Otherwise they will have to use the normal highway, the one that's worked fine until Netflix decided to fill it with lorries.
Those highways between the farm and the neighborhood, though… Anyway, either the analogy breaks or it’s onto something!
Sidebar: I do wish HN had a rule against these kinds of analogies as they do other reddit-esque puns and the likes. The internet is not like a milk truck, it's not like a series of tubes, it's like 1s and 0s being communicated across a worldwide mesh of cables of varying material under the control of varying entities
We're all at least vaguely techy enough that we're on HN, we can understand at least the basics. Lets talk about what it is, not what it's like.
BOB pays ALICE for a connection at fixed uplink/downlink parameters to the public collective of interconnected autonomous networks commonly referred to as INTERNET.
BOB uses the service as advertised to connect to CAROL’s autonomous network.
ALICE fails to adequately peer with CAROL’s autonomous network and calls it a feature.
BOB can’t switch ISPs because ALICE has monopoly on the service where he lives.
ALICE tries to muddy the waters with nonsensical milk lorry analogies that have nothing to do with fiber optic cables to maintain its monopoly and further leverage it to run protection racket on CAROL.
Sounds about right?
When BOB is paying ALICE for the service, he is implicitly paying for whatever “highway” connects his house and CAROL’s milk depots. Everything between the two points is ALICE’s responsibility. If ALICE doesn’t like that BOB mostly orders his milk from CAROL’s then she shouldn’t offer the service as supporting fixed amount of lorries per hour.
It also works the other way around: CAROL does not want to pay ALICE to have premium access.
Also, laughable that you call me a shill. Seems like the most used argument when you don't agree with someone. I'm not even American. So your "BOB can't switch ISPs" does not even apply here.
How does this logically follow? CAROL advertises fixed downlink/uplink connection to any autonomous network. How is it “premium access” to deliver on what you are actually advertising?
It’s like selling SSD drives and then saying oh yeah but if you store video files in this particular video codec they will play at only 15 fps unless the codec vendor pays us extra for a firmware update.
> So your "BOB can't switch ISPs" does not even apply here.
It applies to the particular Netflix/neutrality debate.
As others have mentioned there is no extra "premium" access needed.
Netflix pays for their whole upstream, probably a bit extra for redundant uplinks etc.
Customers pay for the entire downstream.
Everyone in between just have to accept the bits and forward them within reasonable time (Netflix has some caching in the client so it shouldn't be to hard unless someone has oversold their capacity.)
> Also, laughable that you call me a shill.
I looked at your recent comment history and I agree.
That said you really seem to defend an undefendable practice to the point where I understand where people get that idea from.
So I'd rather guess you enjoy annoying people on the internet to see them get mad.
BOB could just switch to using DAVE’s ISP service, who would be happy to connect him with CAROL’s AN without any throttling.
They get paid by customer to deliver bits.
The only extra burdens I see is the "burden" of 1) not double dipping and 2) not overselling their offerings.
All the work to make sure it comply with net neutrality.
not taking extra steps to throttle Netflix,
i.e.
the burden of not double dipping.
I think few people are arguing that Netflix should get special treatment, only that:
* ISPs should treat all bits the same (unless I as the receiver has asked for something else).
* ISPs should have enough capacity to serve their customers.
- basic honesty (not pretending they don't get paid)
- and not extorting money from other companies by keeping said customers hostage
shouldn't be considered a disadvantage?
From oil companies, to fishing companies, to waste disposal to whatever, most companies from a purely technical viewpoint would be more than happy to get rid of regulations that control what and how they can do business. Yet, we have those regulations in place for a reason.
In this case we simply ask that ISPs deliver the bits we paid for. I don't want Comcast treating my bits differently anymore than I want my mail carrier to hold my packages ransom because they seem important and I'd probably pay more for them.
My mail carrier doesn't read my mail. Comcast shouldn't either.
Ie, in a purely technical sense corporations shouldn't want any laws/regulations. But that's not really a relevant fact to anything, is it?
Yes, I think that is the right way to think about things here.
As to the rest of your comments - Imagine if Amazon could pay the USDOT for special lane on the highway.
The Netflix route is using too much bandwidth? Then upgrade it, that's what your customer pay you for. For sure there will be route that will be unequal, some too big, some too small, but that's part of ISP job to make sure its impact is minimal.
Shaking down the destination, because the ISP controls the customer is wrong on many levels.
If Netflix wants to pay extra to deal with saturation they are more than welcome too but if ISPs are denying their subscribers access to a service, even when spare bandwidth was available and sitting unutilized then its time for pitchforks to come out
https://medium.com/netflix-techblog/building-fast-com-4857fe...
The only problem is if the ISP can't actually deliver the bandwidth that they promised in exchange for taking the consumers money. That's called a scam.
What I don't get is, both Netflix and I pay for bandwidth. As far as consumer bandwidth (mine) is concerned, why should Comcast care if it's 1TB/month coming from a thousand sites or just one. I paid for the data, I paid for the bandwidth, give me my bandwidth.
Likewise, Netflix paid for their own too, with whoever is their ISP.
Conceptually the bandwidth has to be paid for, and I could see your argument if I only paid a portion of what it costs to transfer the data.. but that is a broken model if that's the way it is. I want to pay for data, and I shouldn't have to get Movies.com to pay Comcast to serve me movies. I paid Comcast for data, it doesn't matter who it's from. My data is my data.
Is this wrong somehow to you? Honestly it's a strange argument from you, I have a hard time understanding. Like, if you run a website and I visit your website, do you think you should have to pay my ISP for data I'm downloading from your site? That's a bizarre system in my mind.
Netflix (and Youtube and Hulu and all the other big content providers) pay big bucks (millions per month) to get the content from their servers to the internet backbone. Their contracts with their ISPs don't specify where the traffic goes. They just buy lots of 10 gpbs links, and send the data off. They're paying their share.
Your contract with your ISP doesn't distinguish between getting traffic from a mom&pop website or Netflix. It just said they would deliver the rated speed (7 mpbs or 10mbps, or 100mbs) of data to you for your fixed monthly fee.
When the ISPs found out that people were actually using that much data (that YOU PAID FOR), they found they had underprovisioned their network, and couldn't support the load. ISPs started saying, "Netflix is using too much", because, really, it's rude to blame your customer for using the service you provide as contracted.
This is called double-dipping.
So - Netflix is paying their fair share. You are paying your fair share. If your ISP can't handle the traffic demands of their customers, they need to suck it up and provide the capacity that you're paying for.
Edit: I hear arguments like this and I really don't understand them. I would like to hear the reasoning behind it because it just doesn't make sense to me.
I get half the download bandwidth on Fast.com that I get using th FCC Speed Test app, so I don't think the former “only really tells you what the maximum bandwidth your ISP is able to provide you is”.
Once the connection is established you can't see the traffic between you and the URL.
Is my understanding correct?
[0] - https://security.stackexchange.com/questions/86723/why-do-ht...
Even weirder, sometimes this traffic prioritization happens partly in the firmware of the modem the ISP provides.
On one of our test network connections, we can reliably trigger a firmware issue in the router that causes very high packet loss. Then we can "fix" the issue by visiting www.speedtest.net from any computer on the LAN behind the modem.
(My company makes in-browser video calling stuff, so we spend a lot of time worrying about network connectivity.)
And speaking of packet loss, I'd love to see Netflix add a measurement for UDP packet loss at several throughput levels. There's no easy way for non-technical users to do a really good packet loss test.
We often get customer support queries that say something like, "My video calls are terrible but it's not a network issue because I did a speed test and I'm getting 100Mbs." Which is a totally reasonable thing to say. But when we look at call data, the person is experiencing regular spikes of, say, 15% packet loss. That's not something that will effect web browsing noticeably, or even Netflix video very much, but it makes for really frustrating video calls.
Also, there are more links and alternatives under this discussion: https://www.reddit.com/r/sysadmin/comments/5mntko/other_alte...
Ultimately in most cases, I just want to know whether my connection is working or not. If it's working, it'll usually be in a certain ballpark. If it's not (or if I've accidentally left a VPN on), it's obvious.
So for me on the East coast it's not really measuring my internet access, it's measuring my internet access across the whole country.
So using speedof.me I get a result of around 100Mbps.
Using Speedtest.net within my own city it's around 700Mbps.
Using Speakeasy with a Eastern server it's around 400Mbps
speedtest.googlefiber.net to Kansas comes in at around 300Mbps.
So in my opinion the speedtest you should be using is entirely based on who has servers closest to you physically.
I'd love a speedtest that tests to whatever node you're using on all biggest CDNs (CloudFront, CloudFlare, Akamai, Google, etc)
Which many ISP understansibly will refuse to icrease peering capacity and you end up with poor bw speed from netflix or other similar providers. The solution is to do what google does with GGC nodes. Netflix also has similar service but the are very stringent and picky with which ISP they give the servers to. Google basically gives them away to almost whoever asks for it. Not to mention with all streaming and cdn and cloud providers wanting to have their own ggc like hardware at ISP premise it puts cost strains to isp in terms of power usage, rack space and high-output switch and routers at the distribution end. Not to mention all of these hardware will pull bandwidth from source (about 20%) which will come through ISPs expense.
People often like to blame ISPs. While big ISPs have legitimate problems. It also makes it extremely difficult for smaller isps bear this cost and provide service parity compare to ISP giants. Which is why you don't see a lot of isp business popping up. There is a reason google fibre is so limited in scope.
It always facinates me how most ISPs makes money at all. Then i realize the one who do has been doing it for a long time ans were already big companies when started out and probably had to take substantial loans and help from government and even then it probably took them a long time to be profitable.
Where I live there are 100+ ISPs since there's a national fiber network anyone can plug into, so to start an ISP all you need is a router and some peering.
Yes, the last mile is the biggest expense, but other costs are not small either. An ISP with a couple of 1000 users are just reselling BW from someone else. They are at the mercy of their upstream's peering. I wasn't referring to them.
I own an ISP outside the developed nation.
It asks me to disable adblock.
Fast.com has been my go-to site for speed-tests and sometimes for even rudimentary connectivity check. It dethroned speedtest.net as my primary speed-test service for two reasons:
1) I find typing 'fast.com' much easier and hassle-free than typing speedtest.net (aside from the fact that quick '.com' insertion through CTRL+Enter shortcut in Firefox makes it even snappier).
2) Netflix's wide self-run server network gives more uniform results, especially when testing from different continents, because 3rd party server providers -- in the case of Speedtest.net -- have sub-par performance, imo.
This feature is a welcome addition.
Specific case: Our SSH server accessed by users from all around the world. It'd be nice to know the bandwidth & latencies our users get.
pip install speedtest-cli
Looks like Netflix is indeed being throttled here. What's interesting is that we haven't noticed any issues with speed while watching netflix.
Unloaded latency (when there isn't much traffic on your connection through your link to the ISP) generally ranges from 5 msec to 50 msec, depending on what kind of connection you have (fiber, cable, dsl, cellular).
Loaded latency is measured when there is traffic on the link. A good router will minimize the additional latency, but many routers will add a lot - between two and five seconds (2000 to 5000 msec). That's why you lag out in games.
In addition to the new Fast.com test, there are good tools for measuring loaded latency listed on the Bufferbloat page: https://www.bufferbloat.net/projects/bloat/wiki/Tests_for_Bu...
Is there a way to fix this? Seems like there ought to be some sort of traffic shaping measures I could apply at my firewall (I've got a dedicated box running pfSense) to ensure low-latency applications get sufficient bandwidth during upload.
There are also 156KB of images waiting to be compressed to one third of that [1].
Fast.com - is 10Mb normal for 4G mobile connection?
Netflix prefers IPv6. However, they have blacklisted the Hurricane Electric endpoints, because their paying customers with valid accounts might use those endpoints to view the content that they paid for while sitting in another country.
The blacklist does not take effect until you try to actually get video content from Netflix, at which point an error is returned.
The result of all this is that fast.com measures the speed of my IPv6 tunnel to Netflix CDN servers, which will not serve content. I have to turn off IPv6 routing on boxes in my house that want to watch Netflix, and then it falls back to IPv4... a completely different path than the one fast.com has just measured.
With FAST.com you can see if your ISP throttles traffic to Netflix in any way, which is desirable for the ISP but undesirable for you.
Fast.com is just a nice domain to put it on. It could be fast.netflix.com or anything, it just loads the code to kick off the test
If you want to measure the speed you’ll realistically experience in the wild, including your ISP’s peerings and whatnot, that’s another question.
To be fair, I do have about 150 tabs open.