Ask HN: Why is bandwidth so expensive?

6 points by cmorgan8506 ↗ HN
I've been creating video content lately and quickly realized how expensive video hosting ia due to bandwidth.

I'm curious why bandwidth is so expensive.

Also, do you think bandwidth costs are likely to rise or fall in the future?

23 comments

[ 3.6 ms ] story [ 58.5 ms ] thread
An cable long enough to cross the Atlantic cost a lot. But you can also get 10Gbit dedicated line for $1000 /month with ip transit, so I don't think it's too expensive. The more bandwidth you need the cheaper it gets per Mbit. For a startup I suggest going peer-to-peer (like the bittorrent protocol) to distribute your content. Note that you can still charge for it. That's what Spotify did in the startup. With peer-to-peer it uses the user's bandwidth to distribute the content, so that you do not have to pay for the bandwidth.
Peer to peer is an interesting thought. I'm curious how that would work with video hosting. Will have to read into it. Thanks!
Please learn the difference between bandwidth and data transfer.

Bandwidth is a speed, data transfer is an amount of transferred data.

Yes I’m aware that a lot of hosting companies show it as “bandwidth”. That doesn’t make it right.

Bandwidth is like the diameter of the water pipe coming into your house. Data transfer is the like the number of litres/gallons sent down that pipe, and measured by the water meter.

It’s baffling and depressing to me that so many people in this industry get this wrong. It’s like the tech equivalent of the word irony.

Tech people often cannot also tell the difference between megabytes and millibits, MB and mb, so it's not that surprising.
I find that confusion easier to understand. Its often a slightly different prefix or one character case difference.

I think if you always wrote it out as megabytes vs megabits you’d see much fewer incidents of confusion

I appreciate and understand the difference but like you said, most hosting companies use the term bandwidth to describe their data rates. So it makes more sense to discuss it using their nomenclature.

Also, if you want people to listen to you, try being a little less condescending. You can be informative while not sounding like everyone else is dumber than you. Your approach to "educating" makes anyone who didn't know what you are trying to explain feel incompetent. I'm not sure what the benefit of that is for you, but you might want to reassess your approach.

The tech industry is difficult enough to navigate. Why not try and be a little more welcoming?

> most hosting companies use the term bandwidth to describe their data rates. So it makes more sense to discuss it using their nomenclature.

a) if you said to any hosting company, "I need X GB/TB of data transfer allowance" they would know what you mean.

b) given that you're talking about a video hosting site, where data transfer speed is likely a large contender in performance, I'd be surprised if you don't factor actual available bandwidth into the decision making process.

So how do you have a technical discussion about the cost of your IP connectivity, when you're using the same term for the speed and the amount of data?

> try being a little less condescending

You asked about a technical aspect of hosting a web site, using one very specific technical term, to refer to another, related but different aspect of the matter.

In response, I asked you to learn the difference between what you wrote, and what you likely meant, explained what each means, and gave an analogy using plumbing references.

I have zero idea who you are, or what your background is. If you are primarily involved in the business or even the finance side of a company, it's likely you wouldn't know that 'bandwidth' means speed not an amount of data.

> Your approach to "educating" makes anyone who didn't know what you are trying to explain feel incompetent.

Given that you claim to know the term is incorrect, and still use it incorrectly, I accept no responsibility if you feel incompetent.

> The tech industry is difficult enough to navigate.

And deliberately using technical terms incorrectly is the way to fix that, is it?

I'm not debating whether you come off as condescending. I know the answer to that question. Even if you were right, which you are, I don't appreciate your input because of the way you present yourself.

See how that works? You could be the smartest guy on the internet and I still would't listen to you. Because I don't like the way you conduct yourself.

> It’s baffling and depressing to me that so many people in this industry get this wrong. It’s like the tech equivalent of the word irony.

People like you make the industry worse for everyone.

Cheers!

> You could be the smartest guy on the internet and I still would't listen to you.

That's fine, you don't have to listen to me. You've acknowledged that my point is correct, and that you're more worried about how you feel I said something, than the substance of what I said.

Given that you asked a technical question, I'll take "correct" over "warm and fuzzy", Every Fucking Time.

People literally pay me for advice and some then choose to ignore it, why would I care that you choose to ignore it when I offer it for free?

> People like you make the industry worse for everyone.

No, people like me make people like you feel worse, because I follow this crazy idea that words have specific meaning.

It's people who shy away from telling someone they're wrong, or that their idea is fucking awful, that make the industry worse.

> It's people who shy away from telling someone they're wrong, or that their idea is fucking awful, that make the industry worse.

You can tell people they're wrong without being a douche about it. That's all I'm saying.

There was no 'douche'-ness intended in the original comment.

If you've misread my text as having some non-existent emotion or tone, I'm sorry but I can't help that.

Aprt from my mini-rant, data transfer charges vary a lot by host.

AWS for example, have ridiculously high charges.

Can you give some rough figures of how much data and how much you’re paying?

What is expensive?

You can find places like OVH or Online.net that has dedicated servers with "unlimited" bandwidth. Cost is far less than you'd find with the major cloud services. I had a 10Mbps dedicated server for several years on online.net serving 2-3 TB of video per month. Unless you get hammered with many requests, SATA might be enough, while SSD will obviously offer better performance.

Thanks for this. I think maybe I've been barking up the wrong tree with "video hosting" services which appear to charge a significant premium on data transfer.
IP transit is pretty cheap and is expected to continue to fall in the future.

For example here: http://henet.p2knowledgebase.com/view/short_post/WnL8B

You are looking at 1Gbps-on-GigE of $380/mo.

You can run that 24/7 (sometimes they bill you at 95th percentile) but this looks like it is flat rate.

So all the data you can transfer at a 1Gbps pipe.

Jesus, I can get that in Sweden for $100/mo. I've heard of the higher rates in the US, but I didn't expect the difference to be that much.
I've gone on rants about this in the past, but I'll keep this one short: it's only expensive if you're purchasing data transfer from cloud providers and paying per-GB.

For an illustrative example, imagine you have a 1 Gbps (gigabit-per-second) transit line that you are saturating at 95% (~972 Mbps) 24/7. Over a month, you will push 312,075 gigabytes (312 TB) over that line. Amazon transfer pricing is a bit complex to calculate exactly, but for 312 TB you would pay roughly 0.07 per GB for a total of nearly $22,000.

You can rent a 1 Gbps line for under $500 per month.

Think of it like buying a pipe. You can either rent the pipe, or access to the pipe. If you rent the pipe, you're buying IP transit. If you rent access to the pipe, you're buying data transfer. When you rent the pipe, you pay based on its diameter -- the capacity of what you can fit through it. But when you rent access to it, you're charged by how much you put through it.

IMO, this price gouging is the single biggest problem with cloud providers and also leaves a hole in the market. The cloud is pitched as elastic, but it's only elastic so long as your business is not bandwidth constrained. For example, you would not be able to build a competitive CDN or VPN network in the cloud, because you cannot charge your customers for data transfer since your competitors do not charge like that. But you have to pay for their data transfer, even if it all fits within capacity that you could provision for a 40th of the cost.

I've been interested in this market for a long time. A few years ago I was pitching a startup that invovled this idea. It never came to fruition, but I do have an (unfinished) slide deck that might interest someone: https://docs.google.com/presentation/d/1BJICGRhDL95vdBFg0ZE8...

If anyone is working on this problem or wants to discuss it, send me an email as I've got a lot of thoughts on the matter and will be working on a project in this space within the next year.

I think this answers my underlying curiosity. I'm still surprised that no one has come out with a low cost solution that would drive the prices down though. The barrier to entry, which is cost like you mentioned in your deck, must be fairly large.
Exactly. The barrier to entry is not the cost of a single line. It’s the cost of provisioning a global network where each server has a line with sufficient capacity.

If you’re bootstrapping a CDN company, for example, you can’t just start with a single server. You need the global footprint as a selling point. But since you can’t provision that bandwidth elastically, you’re going to have a monthly bill of roughly $1k per POP as soon as you start the company, significantly reducing your runway.

The reason there are so few solutions is because the status quo is charging $/Gb, it’s a huge source of income for cloud providers, and it makes traffic engineering significantly easier because users have incentive to minimize bandwidth costs.

Definitely seems like a market opportunity. Did you find any interest when you were pitching?
I never fully pitched it. That slide deck was mostly to sort through the ideas I had at the time and try to identify some fundamental truths of the market.

The product I'm interested in building is basically a bandwidth marketplace. It's a hard problem and I've been thinking about it for a while. In college I researched "TorCoin" for my senior thesis which explored the possibilities of a proof-of-bandwidth cryptocurrency. I know some ICO's are working on this now but I'm really skeptical the idea can work without a killer product in the hands of users driving adoption. That's the product I want to work on. I'm still figuring out exactly what it is, but I've been thinking about this kind of thing heavily for the past 5 years and feel pretty confident I can execute if given the right time, team and funding.

Really interesting. Thanks for the info!