Hacker News Reader

Comment by vosper | Hacker News Reader

So how much does it cost to store 1gb of data for one month on/with filecoin? I looked all over the site and there was a lot of talk about it being cheap but no prices. Is there some way to compare cost and reliability to S3?

On account of the word “coin” being in the name, I have a feeling the answer isn’t simple.

View on HN

There is virtually no way to really answer that. People quoting the nominal dollar cost are misleading you. Nobody can buy hard drives and keep them running and connected to the Internet for $0.10 per PB per month. The fact that they're willing to offer you that price anyway presumably means they're actually getting paid in some shitcoin they hope will go to the moon. If it doesn't, then you'll find out what it was really costing. The other possibility is the global demand is so low that hard drive owners who already found some other way to fund their capex are just charging peanuts for the slack in otherwise idle disk space, but that is inherently a pricing model that can't scale, like airlines giving away seats on a plane that is not going to meet capacity but needed to fly anyway to keep a route alive.

The answer is: nobody knows, because nobody pays for anything with FileCoin. The services that get pointed to, like estuary, have the price of "free," and most of the incentives in the system are pumped in automatically to "reward suppliers" rather than there being any value transferred between people paying for storage and those providing storage.

Happy to adjust this conclusion - I've spent some hours trying to figure out how to pay for anything with FileCoin and come up empty-handed. Does anyone want to do a walkthrough of buying the coins, installing the software, storing data in exchange for FileCoin, and retrieving it?

In the meantime, you have to wonder: is this a sort of "you get what you pay for" scenario, and you pay nothing, so you get… no reliability?

currently seems to be $0.0000001 USD for 1GB for 30 days https://file.app/

so it costs 10c a month to store 1PB?

My back-of-the-envelope calculations put it closer to $2-$10 per petabyte. The network is most attractive right now for clients who have large amounts (>1Pib) of archival data, although it's also good for acting as a cache-of-last-resort for IPFS. Think Amazon Glacier.

One of the reasons it's so cheap is that this is a quote for literally just raw storage; the other value-added services that you'd expect to accompany that are just beginning to appear now, including the FVM for more complex deals and processing over stored data.

The other big one is an automated retrieval market -- storage providers right now either negotiate separately for uploading and downloading of that amount of data (you mostly have to do that anyway when you're trying to work out how to transfer PiB of data!), or just include it in the price. But that's coming. https://retrieval.market/

All of these will add to the cost, but in optional ways, and not the orders of magnitude that differentiate Filecoin from, e.g., S3.

How do you plan to independently verify storage retrieval? A neutral party? How many?

With the proviso that I am woefully behind on reading up on the retrieval market developments, I understand that there's a few projects being built in parallel with different emphasis. One model is essentially a P2P CDN, where the verification would be distributed across, essentially, IPFS nodes: https://pl-strflt.notion.site/Filecoin-Saturn-efc122f123f344...

Another, which I will try and dig up a link to (if I haven't imagined it entirely) is more of a sentinel system, whereby clients and storage providers have a protocol to pick a neutral third party that escrows and streams FIL payments from clients to SPs, conditional on retrievability. But as I say, I'm not an expert here.

Apparently that does not include data redundancy. You must take care of that yourself by buying the same storage from multiple vendors: https://docs.filecoin.io/about-filecoin/faq/#after-i-made-a-...

Also, the next FAQ question answers an interesting problem where the cost of data retrieval is not agreed before hand, and allows the storage providers to extort you for actually ever retrieving your data. Solution is again storing your bytes with multiple providers and hoping they are not all extortionists.

> Apparently that does not include data redundancy

Indeed. So if you compare it to S3, the calculation seems to be "Filecoin cost / Amazon cost = 0.0010% the cost of Amazon S3", so if you're willing to pay the same amount as for S3, you can store your data with 100,000 different providers around the world.

I'm not sure how many locations S3 data get stored at, but I'm fairly certain it's not 100,000 different machines, so even with that, it seems like a pretty good deal for vital data.

If you want it cheaper than S3, you can store it with just 10,000 different providers (10% of the price of S3), and I'd still consider it a good deal.

The problem here is end-hosting transparency. How Amazon and S3 and the other providers structure themselves to ensure redundancy is fairly well known.

How those 10,000 providers do is quite different - namely, a 6000 of them are all behind the same segment of the internet that can go down? Then if you were distributing data aggressively, you probably lost access to something temporarily - or permanently.

This isn't an idle risk: if it's cheap and easy to start making some amount of money off of Filecoin hosting, then you have the fly-by-night problem: suppose some Chinese firms with cheap warehousing cobble together a whole lot of storage, and then, to optimize against people doing exactly this (and thus increase revenue) they advertise themselves out onto the network as dozens, or hundreds of logically separate entities?

What looks like cheap storage which you've made redundant, suddenly actually isn't - it all has the same base fault possible.

Interestingly, this is PRECISELY what FVM will help enable.

By using an FVM contract, you can specify exactly what you'd like - how distributed, how automatic the replication, etc etc. This, of course, is a bit manual for now, expect lots more simplicity in the space - or just use https://web3.storage which does this all for you!

Disclosure: I am co-director of Research Development at Protocol Labs.

A few things work in the network's favor in this situation (disclosure: I work at the Filecoin Foundation: I guess David and I are tag-teaming this). Firstly, the proofs that storage providers have to provide to the network are designed to work per replication. So you can't store just one copy, and claim you're multiple providers storing multiple copies.

Secondly: storage providers aren't anonymous, and in fact compete on additional features on top of the commodity storage offer. At the most basic level, you have orgs competing on geographical diversity -- so you have PikNik https://www.piknik.com/ or https://www.sealstorage.io/ in the US, dcent in the EU https://dcent.nl/ . And then there are other contractual requirements that let storage providers offer distinctive services, like the use of renewable energy.

Thirdly: the economic requirements of the Filecoin proofs at this point (constant availability, a reasonable amount of CPU/GPU power, an initial FIL stake that can be removed in the event that you fail to prove in a timely fashion you're storing the data), mean that the optimum storage provider setup is pretty professional -- you really need a presence and fast connection in an extremely reliable hosting center, a 100TB or more, and so on.

So, I'd say that there's a (potentially contingent -- i can imagine see the entry price for filecoin storage provision going down) entry barrier, and a lot of meta-data that can, and is, being used to make decision about which SP's to make deals with. So, for instance, Estuary and I believe the other front-ends use historical and business data to determine who they're making deals with.

What is the value FileCoin is providing in this scenario though? I feel like there's a bait and switch here: proposition 1 is: FileCoin lubricates storage provider transactions by providing proof of storage, such that you can easily transact with a gigantic marketplace of vendors.

But then that's affected heavily by: to ensure storage is secure and redundant, you'll need to research extensively which providers you contract with based on reputation.

Which pretty much completely undermines proposition 1: I could just transact with providers who act as registered business entities in countries who's legal systems I trust, using regular currency directly. Because that stance in the marketplace is a shortcut for "is a geographically distributed operation that can be held legally accountable for their behavior", and have a history of success backed in the regular market place.

Pin on IPFS with vendors using totally different networks. There is also Sia and Storj.

This is broadly correct! It's remarkably cheap. As other commenters have noted, you'll want to handle reproduction/etc, which is a bit more work, but still many orders of magnitude cheaper than anything else out there.

HOWEVER, most folks don't want to do this - we recommend solutions like https://web3.storage (which gives you 1 TB with three replicas for free forever), and https://estuary.tech.

Disclosure: I am co-director of Research Development at Protocol Labs.

> many orders of magnitude cheaper than anything else out there

How does this work? Are other providers earning many orders of magnitude in profit, or are filecoin's storage costs being subsidized somehow?

It seems to me that, at some level, they must be subsidized by new buyers of filecoin. It's hard to imagine anyone providing storage so cheaply even if they didn't make any profit at all. Miner revenues have to come from somewhere.

What I want to know is whether a good equilibrium is available. Say some decentralized apps with significant storage requirements win big over the next five years, and the filecoin network starts seeing a more moderate level of usage. Assume also that sentiment normalizes so that the filecoin market starts viewing filecoin as a service token for storage and retrieval, not as a speculative investment. Is the network still competitive with centralized storage providers, and how does the filecoin supply inflate over time as the network operates?

So the initial capital investment is supported by the value of the coin, and that's what gets the network to its current capacity of c.17 EiB (https://dashboard.starboard.ventures/dashboard ) . Committing storage to the network gets increases the probability of gaining new filecoin, so there's always an incentive there to grow the network's size, and once you have that storage available, there's an additional incentive to seek out deals to use the space.

I'm not an economist, so I can't speak to the details of the balancing act, but the incentive system is engineered to create a balanced equilibrium -- so more demand increases the incentives to provide storage, etc. It's a challenge to follow all the thinking, but this video from ZX Zhang is a good guide to the current state (I fast-forwarded it to the relevant bits: https://youtu.be/gbJgsav2lP0?t=554 ). There's also one of the regular cryptoecon days coming up in Paris, and the videos of those are kept online. https://www.cryptoeconday.io/event-schedule/cryptoeconday-et... https://www.cryptoeconday.io/videos

The network doesn't exist in a vacuum and that 17EiB isn't free. People had to buy real hard drives using dollars (or whatever) and they have ongoing opex to keep them spinning and connected to the network. Those dollars come from somewhere. If you're getting storage from the network for less than the amortized capex and opex of those drives then it's being subsidized, either by the owners of the hardware (if they're mining at a loss) or by people buying filecoin for purposes other than paying for storage.

Two sources of storage provider revenues are block rewards (minting) and deal fees (paid between storage clients to storage providers). Indeed, there are real fiat hardware costs to providing storage/related services (cost of hardware, computation etc). Given that the network is fairly new, we expect that currently storage provider revenues come primarily from block rewards, and deal fees are minimal (as you noted, storage is cheap). In short, the network is subsidizing capacity onboarding through block rewards; FIL as a utility token has some intrinsic value (storage providers would not incur fiat operational expenses if they did not receive commensurate rewards in FIL). As the network matures and becomes more robust and trusted, rewards from block reward minting will naturally taper, and storage deal fees will constitute a growing share of storage provider revenues.

Regarding the question of a “good equilibrium” and the ability for the network to “compete” with centralized storage/data services, we believe that as the network matures, and quality of service improves, it can certainly be a viable competitive option, with additional unique use cases separate from the current offerings from centralized players. The ability for clients and storage providers to come to market equilibrium prices for customized storage, retrieval, and computational services addresses the inherent inefficiencies in economies dominated by large, centralized players with limited choice.

How the network’s supply evolves over time is subject to many exogenous factors, but, in general as the network matures, we should expect the economy to be more deflationary rather than inflationary, as token vesting schedules end, and block rewards (minting) taper as mentioned earlier. Please see here (https://www.youtube.com/watch?v=mj4QxYKPMmA) for a discussion on Filecoin’s circulating supply, and the sources of inflow/outflow that determine its dynamics.

> for free forever

Translation: Free for maybe about 5 years if the company doesn't fold or less if it does

I think it's currenlty free? https://web3.storage/

Take a look at web3.storage and Filebase for IPFS pinning. Pinata if you know it will be a small amount of if you are rich.

https://file.app/

Pricing for web3 projects is a bit more complicated than getting a product from AWS. As you participate in a specialized economy it is like asking what a Danish stock costs from the US standpoint. First you need to buy DKK, and thereafter the stock might be volatile and have another price. (Spoiler alert, Danish stocks are cheap now :) )

Reply on news.ycombinator.com