Longterm business viability of storj

I have two questions regarding storjs long term viability

  1. Do you really think that relying on non professional storage will longterm provide a stable business model. The question arises from the fact that providing storage to storj is only viable when it is a hobby and not really serious. I.e. it is not done to provide storage with the full intention to keep it available, but instead you had some leftover and don’t really care. At some point the demand will exceed what this class of storage can offer.

  2. Untitled spreadsheet - Google Sheets

I have put the cost together and had to see that the cost model is actually cost negative, let me know if you know if there was a mistake, but otherwise my question is: Is storj gonna raise the cost in the future and this is only a promotion price?

1 Like

Yes. The moment Storj actually starts needing exabytes of storage, it will be viable for professional actors to contribute storage as well. Consider that right now Hetzner sells non-redundant storage at ~1.2 USD/TB. They’d probably be quite happy to host Storj data at 1.5 USD/TB the moment Storj actually starts having enough data for professionals to care.

See New pricing announcement for discussion on this topic.

Yes. The moment Storj actually starts needing exabytes of storage, it will be viable for professional actors to contribute storage as well. Consider that right now Hetzner sells non-redundant storage at ~1.2 USD/TB. They’d probably be quite happy to host Storj data at 1.5 USD/TB the moment Storj actually starts having enough data for professionals to care.

Do you have the source for that?

1 Like

thx, but you didn’t realize that this is a misconception.

They’re not selling at 1.2USD/TB/m, they are selling 76,75€/64TB/m. This is a big difference. Storj can’t offer to immediately take 64TB, so Hetzner would never opt to place that machine in b/c it would generate cost instead of revenue. Plus, in the current model it is discentivized to run on raid, so assuming one of the disk fails the electricity and location was still payed for that time, but the money is gone. The held back model of storj doesn’t work for the real world, where professionals make guarantees upfront. In multiple dimensions it would be a bad decision to do it. As a short list:

  • money is held back and not guaranteed
  • availability is discentivized and delivers negative return
  • the held back money effectively translates into a credit towards storj. Storj does not pay the market rate of 7% per annum for the additional missing revenue
  • you get only a funny joke of money during the first 24 month (which is already the majority of the lifetime of the server running the disk) of what you should get immediate

It has been worth it for me, as a hobby, but I also have plenty of spare disk space.
The held back payment is insignificant after two years imo

It’s not designed for pro SNO, that’s true. We always ask to use what is already online, so costs already paid.
Held amount is negligible long run as correctly mentioned by @andrew2.hart, but you of course right too, if you start with a new hardware and make it online only for Storj - it likely will not worth it, because it will generate costs at start instead of revenue. It may generate profit after a long run though, but it’s highly depends on your costs (not an advise), see Cost of running nodes?
The second thought - pro (expensive) hardware is not required, storagenode can be run even on the router: Running node on OpenWRT router?, using a pro hardware will not give any preferences in Storj network.

So, at the moment using non professional storage is a benefit. It can be so even for professional hardware, if it’s used not only for Storj, and you offer a not used space to the network. In this case, all costs have already been paid and all income is net profit, this will be a discount to your bills.

1 Like

The thing is I am a professional that is why I am speaking from this side and storj got my interest but lost it fast.

Held amount is negligible long run as correctly

Yes, for a hobby it’s okay and a good mechanism to force people to stay on the line. But as said before, with this move, Storj effectively asks for a credit from the provider which it does not pay for. So should Storj at some point really consider making it attractive for professionals, it will need a secondary model.

It can be so even for professional hardware if it’s used not only for Storj, and you offer a not used space to the network.

Hardly, the reality is you rarely have so much storage floating around forever. It is more of a moving window. Plus those disks are usually already in a RAID, which is disincentivized by Storj.
Since it is a moving window, that space is not free forever but needs to be made free again. So in the best case, you have a 34 LFF Chassy and only need 4 of them at the start and fill the rest up with capacity already and use it for Storj. Now after a year, you will need that capacity, but it’s neither worth replacing the capacity to continue running Storj on it nor was it worth running it at all b/c the storage anyways would not have been filled up due to your IP filtering.

I really like the Idea behind Storj and I would have liked to support it more, but obviously, I won’t b/c it just makes no sense. Also, Storj is not a fit for high-performance classes of consumed storage, it is only good for slower purposes, so while video streaming is viable due to the sheer mass of individual uplinks, model storage for an AI or physics simulations is not.

My questions, in the end, were honestly targeted at getting an answer: Is this something that will change in the long term, or not?

Which is why I added the qualification of:

Filling a box of 64 TB will be easy when Storj starts holding exabytes. This is not the case yet, so if you are a professional wanting to spend professional time on a professional SN service, this is not the time yet. I know I’d love to, so I’m waiting for this to happen as well. For now my storage node pays for my electricity bills, deprecation of hardware and leaves a bit for buying new board games, so I’m happy with the things as they are now.

Besides, if exabytes of data do not happen, then there won’t be enough long-term funding for development of the software, so it’s quite an important condition for Storj as a whole to happen.

Also, the margin for Hetzner might be even better with: (1) not having to deal with those pesky customers, because Storj already does it for them, (2) using boxes designed for Storj, and not just off-the-shelf dedicated machines. The 0.3 USD/TB margin is the lower bound of what a capable professional can do.

BTW, I am not sure if you are aware of that, but storage is a very low-margin business. It’s quite difficult to earn a living by just offering storage. Storage services earn their margins on value added on top of storage, not pure storage, and by being a SNO there’s not many ways you can add the value to your offer. So, doubly so, maybe being a SNO is simply just not for you and your skills can be put into better enterprises. There’s nothing wrong with that.

Filling a box of 64 TB will be easy when Storj starts holding exabytes.

No it wont. B/c of competition, even if it holds exabytes (which is a small number btw.) upcoming competition by the chances arising would just regulate itself. Without a model that can aquire a larger quantity of storage at once this wont work.

Also, the margin for Hetzner might be even better with: (1) not having to deal with those pesky customers, because Storj already does it for them, (2) using boxes designed for Storj, and not just off-the-shelf dedicated machines. The 0.3 USD/TB margin is the lower bound of what a capable professional can do.

I know how low I can drive the cost of storage per disk, and your numbers are not correct, they can be smaller though. However, I also know internals of market competitors including Hetzner. And their margins are not as big as you think for those machines.

BTW, I am not sure if you are aware of that, but storage is a very low-margin business. It’s quite difficult to earn a living by just offering storage.

Not really difficult no. That is just simply untrue. Customers are willing to pay for a secure storage service a lot if they fulfill all the requirements. I can tell you the prices of Backblaze and AWS s3 and even Azures Archive which has the fairest pricing on the whole market (compared to directly publicly available services, there are business only services that are less expensive but are usually marketed by direct targeting), are not low margin. It depends however on what you compare. Of course if you compare against a full solution with compute, the margin looks small, but compared to other businesses its big (ranging between 30 to 100%).

You can say the same about filling a single 16TB drive, you won’t fill it immediatelly. And we know large storage services do use them.

If so, I am ignorant of your knowledge. Care to share, so that we can discuss it?

«secure» is also value added. SNO aren’t supposed to operate secure or reliable storage, and so Storj won’t pay a premium for more security or reliability than bare minimum. So does maintaining APIs more complex than a key-value store, which is what SN basically is.

The spreadsheet is hard to decode and doesn’t contain a chart showing a break-even point. With a guaranteed (for the time being) $1.5/TB/month (and assuming a RAID setup to prevent loss of the terabytes stored over the years in case of an HDD failure) it is obvious that in the Storj’s payout model the break-even point exists - so please put such a chart in the spreadsheet. Otherwise, in my opinion, the cost analysis cannot be taken seriously.

you can put it yourself if you want.

This chart is not for the operator and when they turn profitable, it is for storjs business model. The final cost does not consider the raid, it just listed it for comparison.
The fact is, Storj currently pays for every TB 1,74$ (at least, not including labour cost and other factors) extra that the customer doesn’t pay to them. So you want to know the break even? Increase the price by 1,74$.

You can say the same about filling a single 16TB drive, you won’t fill it immediatelly. And we know large storage services do use them.

Yes, but they buy them b/c they exactly know how much increase they will have for the time period they’re looking to use them or how much the customer booked. If the customer doesn’t book the full disk, then naturally the price goes higher, this is how economics work in general. If you take bulk you get discounts.

If so, I am ignorant of your knowledge. Care to share, so that we can discuss it?

Never said or intended to say you’re ignorant. 0.4 is the average low bar currently for disk prices in retail, if you insure yourself instead of leaving it the manufacturer that can be lowered again, with that we are already at (or below, but rarely currently) the 0.3, depending on the conditions you get. The rest you can’t save on the hardware level, but in the design of how you serve the storage, how you tier it and most importantly how you post utilize the drives post 36 month, usually they should be part of the design.

My perspective is a little different but i share it anyway:

I become a node operator a lot of years ago, at the time storj was not even called storj but SJCX.

With my work I am able to get wholesale computer components at very low cost, so I thought about building nodes specifically for storj. So I drew the case in 3D, printed it and finally assembled it.
The first node was a success right from the start, at the time there was no barrier on the new nodes and the stress test.
So I thought about expanding the operation, and since again, for work, I have access to different networks, I talked to the various managers and in exchange for a discount on my work they allowed me to put the nodes on their networks. In this way my work cost something less than the various competitors, and on the other hand the economic part that I gave up would be returned through storj. That was another big success for me, the new business model works and my own customer ask for it!!!

Few years ago, there was a spike in price, so i sold almost all my storj holding, and use a part of this to replace all harddrive with a bigger one. In such way i had more space to fill and brand new drives (and income).

Last year there was another spike, so i sold another big amount around 2$, but considering the 2$ it bring me a lot of money, and ask to myself: “is not better to expand further my node network instead of double the disk again?”

So, last year i prepare and put online more nodes…today i can say that it was a terrible idea.

While this nodes give more network power on storj network, more resilent than just have one single big node and they are powered by different power grid…they are unused because the “new node” barrier. So today i have the good old nodes filled, on about 4year old drives (with can put them in danger) that make me a total of about 400$ per month and the new nodes almost unused for no reason (at least from my side) to which I will still have to change the discs in the years to come.

So, I tried to upgrade one of the old nodes, to understand if it was the network full or if it depended solely on the “new node” factor. Well, the old node last month (the first after the power up) increased the used space by about 300M while the new nodes put online 8month ago still gain something like 100M (each one) … WTF??? Never again, i will upgrade all old nodes very soon and i will double them again in next years until the deletation rate will not allow them to rise more.

While i understand that my situation is unique, and the network is build for every one case…i had one question:

In the current way, first or later, all old operator, me included, will have something like 50TB or more on the same network (some one of them already got it), but their network is a resident with lower upload…what happens if the node start to be used for real?
YES, there are a lot of node…and the data can be download from different place BUT what if a lot of nodes are in such situation? Currently there is no big data usage and the only limit are the success audit, but if the network become heavily used what happens? A big part of this network start to fail the audit (that today maybe success just because no other usage is in place), got disqualified and leave…and this will not be even worst for the network effect?
YES, the grace-exit and the fail audit will “protect” you from data loss (maybe), but not from network distruption.

For this reason i think @wzrdtales have right, currently storj is usable only for slow purposes (backup mostly), because if some big company (data center like) try to use storj it will put the network almost offline (that is stupid considering the decentralization).
The storj network work only if a lot of space and bandwidth are free, otherwise the whole network collapse.

A dual model system is a possible solution…so, on one side you keep the decentralization, allow to reuse old hardware, hobbist, home connections etc etc…on the other side the professional keep the network safe allowing storj to be used even by big coops.

it is more slow b/c storj does not incentivize fast disk and doesn’t categorize either, rather than slow upload. Slow upload would be a show stopper for almost any kind of storage, but I doubt that that isn’t something storj doesn’t already monitor.

Also its unclear if storj can actually achieve read-after-write strong consistency.

Slow nodes will be canceled due “to long tail cut”, so you would upload to the fastest nodes and download from the fastest nodes (110 nodes will be selected for upload and uploads will be cancelled when the first 80 are finished, and 39 nodes will be selected for download and downloads will be cancelled when the first 29 are finished - you need to have only 29 pieces from 80 to reconstruct a segment).

The Storj network has achieved strong read-after-write consistency. Your object will not be available until you have finished uploading.

The Storj network has achieved strong read-after-write consistency. Your object will not be available until you have finished uploading.

Useful to know thanks for elaborating.