Internet connection. Why is download speed much faster than upload?

I visited The ping is 32ms. Download speed is 5.07Mbps. Upload speed is 0.63Mbps. Why the huge difference? What is the constraint?

I think ISPs just set it that way to save bandwidth.Most people (ordinary users) download far, far more data than they upload, so they do not notice the relatively slow upload speeds anyway. It is the download speeds that get advertised and over which ISPs compete (when they do). It is assumed, usually rightly, that customers do not care about upload speed.

Most people consume way more than they produce.
So, the bandwidth is allocated to reflect that.

Wow! That’s slow… I just checked mine at - Ping 30.13, Down 37.53Mbps, up 8.75Mbps.

Now I am not on a straight fibre connection as the last 100 metres or so are through a twisted pair. It would be interesting to see other people’s speeds.

Faster uploads are generally only needed for enterprise(business)-level internet connections. They will sell you a connection with better upload, but it will cost more money. One of my former roommates operated a side business out of her home as a wedding photographer. She would upload full-size proofs online (many gigabytes at a time), but was too cheap to pay for a business internet connection. On those days, I would be totally unable to do anything online, because her uploads sucked up the entire connection.

I’m in the UK and I get at home:

75 Mbps down
15 Mbps up

The A in ADSL highlights this, and was made that way intentionally as said.

:o yeah… I’m not posting mine. Needless to say, if I want to compare to the speeds posted, scientific notation would be helpful. :frowning:

The average consumer user of the Internet does so in a browsing fashion. Typical browsing is a tiny request up (“give me a web page”) followed by a flood of data down (“Here it is, along with 1500 ads you don’t want”).

This allows the providers to promote the downward speed and ignore the upward speed. Rarely is the upward speed mentioned unless you ask, and the agents often don’t know what you are talking about when you do (it’s not on their info cheat sheet).

In the past, running a web server from home was frowned upon, since it might use more bandwidth that a typical web surfer did. Pricing was developed so web servers would have to pay more. This may be less of a problem now, but momentum is a powerful force, even in economics.

This allows ISPs to advertise rates that are unrealistic to many, but maximizes their income as long as no one presses too hard.

Traditionally (“in the old days”) up and down was symmetrical, in the sense that they were the same and both sucked. :smiley: You would either have some pathetic speed on a dialup modem, or pay an unbelievable fortune for a DS1 (T1) to give you 1.544 Mbps up and down, which almost no consumer could afford to do even if it had been available in residential areas.

The asymmetry began, AFAIK, when cable and DSL more or less simultaneously started offering consumer broadband. The bandwidth requirements were now so high that considerations of efficiency made it desirable to divide the up and down bandwidth in ways that reflected actual consumer usage. For cable, offering excessive (and typically unneeded) upload bandwidth to residential users would impact the bandwidth available for television and other services, including digital telephone service over cable which happens to sit (in the implementations I know of) right in the upload part of the spectrum. From the ISP’s point of view, a bonus side effect for them is that the lower upload discourages the operation of a server on a residential service.

What are you paying for?

My subscription is asymmetric, which reflects my use and most users, and allows the provider to maximise the use of the available bandwidth.

Very true. But I wonder if this is due for a change soon. A web server is not all that uncommon anymore; some webcams can operate as such, and the trend is towards decentralization of data, not concentration.

Nevertheless, I think the ISPs and cable companies will milk it for as long as they can. I happen to have it on good, inside information, that if my cable supplier were to allow everyone on my feed 1000Mb/s in both directions, that the infrastructure wouldn’t have to be upgraded one bit and wouldn’t cost them a penny more. (Fiber has an amazing maximum capacity for an incredible low cost.) It’s only for greedy, and IMHO, misguided economic reasons, that this doesn’t happen.

True, and I will occasionally run a web or FTP server myself for some specific and usually short-term purpose rarely involving much bandwidth. Most ISPs haven’t dared actually block those protocols on the upstream side because they are indeed important data-sharing mechanisms. What they don’t want is inexpensive residential accounts being used for commercial servers.

Ha ha – yep – in a word association game, if someone says “cable company” I think “greed”! Also “incompetence”. :smiley:

Trouble is, though, that although what you say may be true for fiber, fiber to the home is still relatively uncommon, and saturation of neighborhood cable segments is still a huge problem with cable broadband. Not that DSL doesn’t have its own issues.

My neighborhood is served by fiber to the hub, distributed by coax from the hub to 32 residences max. One fiber line serves about 8 hubs right now. (This info comes from reliable sources, but could be obsolete, and could be wrong.)

My case may be unusual, but I have never been able to detect congestion, which is not surprising, since most of my neighbors aren’t here 6 months of the year, and when they are, it’s grandma getting emailed pictures of grandkids that make the bulk of the traffic. Very few watch videos of any kind, let alone hi-def. Not many know how to spell or even find YouTube.

There is a 3" plastic pipe underground that houses the active fiber. I suspect that if the entire 3 inches were filled with fiber, we could carry much of the world’s traffic past my house; fiber has such an amazing capacity, and fiber is far cheaper, smaller and lighter than copper wire.

In short, policies set in St. Louis, 600 miles away, are not appropriate for all neighborhoods served by the same cable company.

The reason that ADSL is used is technical, not just an ISP choice. It comes down to issues of signal to noise and crosstalk. This is a peculiarity of using bundles of twisted pairs (aka copper phones lines).

The useful data rate you can get is limited by the signal to noise the receiver sees and the bandwidth. The more noise there is on the line the worse things get, and the data rate drops. The dominant form of noise is crosstalk from other lines in the bundle. Complicating matters, DSL pushes the bandwidth available as hard as it can, so much so that attenuation (ie signal loss) as you go down the wire to the receiver is large, and is the other limiting factor. But it also brings about a curious asymmetry in the system.

All the cable bundles are concentrated together at the DSLAM (the ISP’s equipment) end. But as the bundles fan out to the customers, the customer end is very sparse, and the bundle in your street only contains the lines for houses in your street. At the DSLAM end there is a huge cacophony as the DSLAM sends all the download signal out to all the customers. This unavoidably crosstalks between lines. However if you measure the signal to noise of the signal coming out from the DSLAM it is still pretty good, because the signal is at its maximum amplitude. As the signal travels down the line to the customer, both the signal and the cross talk noise are being attenuated by the same fraction. The line does continue to pick up noise from nearby signals, so the noise continues to rise, but because the lines are fanning out everywhere, this effect is reducing all the time as the number of nearby lines reduces. In the end you see a reasonable signal to noise.

However sending a signal from the customer to the DSLAM isn’t so lucky. You start out with a nice clean signal, but it is attenuated on its journey to the DSLAM, only to arrive there at its weakest, to be presented with the huge cacophony of crosstalk noise at full tilt. Thus the signal to noise for a signal arriving at the DSLAM from the customer is much worse than the signal to noise the customer sees coming from the DSLAM.

So, intrinsically the available data rates are asymmetric. You can perform some tradeoffs. If you devote more time to the customer to DLSAM link, by reducing the time on downloads, you can improve the upload data rate. But you pay an asymmetric cost. You will trade off over 3 times as much download data to get an improvement in upload. Some ISPs provide a range of options, and some allow you to choose “Annex-M” as a profile. This can up to double your upload rate (maxing out at 2Mb/s) but it will chew up a lot of download. Note that if you are already at the end of a long or noisy line, swapping profiles won’t give you such big changes, as the higher frequencies needed to get the high data rates are probably not useful to you no matter what you do. But for high quality links, close to the DSLAM, the tradeoff is very real.

It started on dial-up, in fact. So-called 56k modems had ~53k download speeds but ~33k upload. Some versions of the 56k standard allowed more balanced speeds, but I believe the asymmetrical version was more popular.