By my (very very very) rough estimates the value of the resource circa 2020 is :
$80 per Mbps-year for bandwidth.
$100 per Terabyte-year for storage (for a 10x min chunk redundancy)
Presuming the store-forever storage cost is (again very) roughly 4x present storage cost gives you a value of $0.0005 per 1MB chunk delivered by a GET. That is what farmers need to earn to break even. So I guess you could call that $0.50 per GB stored (forever).
Clearly the logic is off somewhere considering how low the filecoin price is by comparison.
Maybe the logic above is fine. I forgot that the filecoin erasure code is 5x more space efficient than 10x replication.
Factor in loss of purchasing power due to inflation and the real cost (roughly) approaches 4x. Joking aside, Good to see additional confirmation that forever storage cost is no more than 5x year one cost.
Filecoin has 39.9K Twitter followers
MaidSafe (which barely posts) has 51.6K
More people know about MaidSafe than one would think. At this time, the issue is less about awareness and more about belief in MaidSafe’s ability to deliver a working product.
On principle, I agree with you, but corporate inertia is a terribly difficult thing to overcome, even when there’s an obviously sound strategic choice.
Interesting and obvious when you say it like that… the long term saving, tempts those interested in long term data storage… more than where data turns over quickly. Static data applications then likely more inclined but offset perhaps with a need to be confident in stability. All bodes well for a balanced takeoff over time… but unclear how the number of coin play into that.
I don’t know how Filecoin works. if price based on supply and demand? Right now they have an enormous capacity with peta bytes on peta bytes of supply and probably low demand on storage. It would make sense if the price is well below brake even. but do not know if mining rewards offset the low store price but that would mean an unsustainable long term model unless endless printing of coins or future demand improves.
When did some research on IPFS a few years ago the conclusion from listening to others where that it was bad code, speed to finish over quality. Basically that the code was trash, full of security vaulnerabilities and need full re-write. A teacher told me once that shit input equal shit output, because of that I have not bothered about Filecoin or IPFS.
And when you reported of what hardware was necessary to run a node it just confirmed for me previous conclusions, shit code, shit input equal shit output / results, speed to finish over quality.
Disclaimer my knowledge about code and technology is limited but probably above average.
Thanks for making the depreciation explicit. I was implying that with Arweave’s pricing on the assumption that, knowing the group that funded them, they likely took into account the cost of storage over the useful life and reverse amortized it in setting their price range.
There’s a mistake in my $3 - $8 estimate above. Reading this post too quickly, I took 0.276/GB to be per month. But it is per year.
Let’s go to the source:
S3 standard at 50TB costs $0.023/GB/month
S3 Glacier cost $0.004/GB/month
Assuming that storage on Safe would be a mixture of these types of storage, let’s average these them, which yields $0.0135/GB/month or $0.162/GB/year (a more thorough approach might average all the s3 pricing tiers).
Hence:
Lifetime cost = 10x - 30x monthly cost => $0.135 - $0.405/GB. A far cry from $3 - $8 borne of the $0.276/GB/month error.
Lifetime cost = 60x monthly cost per your depreciation calculation => $0.81/GB.
Did you find how to use it, upload files and such? Was thinking about trying out IPFS and compare it with Microsoft, Amazon cloud or similar. Thinking about backup a few gigs of important data and comparing the services and just for privacy throw them in an encrypted folder, because they are backups so don’t need them unless a solar flare or similar happens.
Went visiting the IPFS page, it was easy to find how to download software to become a miner but it was not obvious how to download and such if you wanted to use the service.
Each of CPU / GPU / RAM / Storage plays a part at various stages of mining but never all parts needed at once.
It’s not about erasure coding.
No, but only for lack of motivation. If other people have to pay to download what I’ve uploaded then I’m not interested since I can’t easily share my content with people.
I normally check https://filstats.com/ a couple of times a day to see how they’re tracking. Usually filecoin storage is between 5-20x cheaper than S3 and about 1000-5000 GB uploaded in the last 2.5K deals.
Just now I’ve seen that filecoin is more expensive than S3 for the first time - 1.79x costlier. This is also clearly reflected in only 17.501 GB uploaded in the last 2.5K deals. That’s a massive drop in uploading.
The strange part is FIL is currently valued at about $28, not much different to the last few weeks. So why is storage suddenly so expensive?
As an aside, network storage is currently at 994 PB, pretty huge.
That is quite big risk for anyone to build something with this quickly changing prices:
Cost vs Amazon: 2.16x Costlier
Average Deal Size 0,007 GB
Total FIL paid (last 2,500 deals): 0,168 FIL
Cost per GB (varies by hour): $0.597 USD
Edit: 2h later:
Cost vs Amazon: 1.66x Cheaper
Average Deal Size: 0,018 GB
Total FIL paid (last 2,500 deals): 0,114 FIL
Cost per GB (varies by hour): $0.166 USD
I agree and probably add a point. If the “farmers” “miners” “earners” etc. have a significant investment in hardware to recoup this could matter a lot when it comes to minimum price they can accept to stay investing in more hardware. (which should reflect to end user resource costs in a balanced supply/demand cycle)
So the base costs matters. If the base cost is small or even negligible (use existing free space on commodity hardware) then the “tolerance” to price fluctuation is easier to deal with.
Safe since day 1 has targeted the low or zero cost base and I think it’s under-looked a lot, but is really important.
I am not trumpeting some “we were right all along”, for a start we have not launched, but this is where I see fundamentals matter. I also don’t know yet if our “earners” will always be low resource requirements, but that target is very important to maintain. It’s too easy for software devs to say “let them eat cake” and by that I mean “force them to buy better computers” and this is wrong.
It’s why we keep looking at Arm cross-compiling and so on, small devices matter, cheap devices matter and removing base costs are likely much more important than folk realise. This is one area I watch with interest, were we wrong? should the nodes be supercomputers in their own right or not, I am still of the belief that they should not.
Operators operate a node. Sometimes they earn, and become earners… no guarantees though. Regardless everyone wants their nodes to operate efficiently without churn, so they must be good operators for this to be true by definition. Let’s also not forget the importance of good math operatators when dealing with computer software
Joking aside, Good to see additional confirmation that forever storage cost is no more than 5x year one cost.
