They are reducing it as mentioned in the top post. Just gradually as to not cause shock and SNO exodus, which would be bad.
Here’s my question. How does this cost develop over time? Storj is small scale and can currently hide among the vast majority of consumer ISP connections without disrupting the pricing models in place. Consumer ISPs have always had customers that lose them money, they just make sure they make money on average. If Storj would keep their payouts stable at $5 per TB egress, by what time is that likely to become feasible in the network backbone markets? In short, is it possible for them to ride it out at that payout level?
Now it may be unfair to ask you that as it makes the operation you run likely lose money over the short term and that sucks. But it seems to me that it may be possible if Storj scales responsibly to be ahead of cost drops and stay afloat on overused consumer connections until this level of pay becomes more reasonable. Am I wrong?
Not necessarily as growth from customers can compensate for the loss. And at the least spreading the impact will ensure SNOs don’t all leave at the same time.
It depends on a number of factors that are currently unknown. Just how much the cuts are and whether the data is downloaded more often by a significant margin. If it is mainly backups - as you admitted you use Storj for yourself - no egress payments. If SNO’s only earn around 40% of current levels and old nodes can’t expand above their previous size levels the data that replaces the synthetic data would have to be quite active to fully compensate.
Bandwidth cost has constantly been increasing since the dotcom bubble burst in early 2000s with modern internet infrastructures deployment. This has also been made worse as we are running out of ipv4 ranges. The battle for IPs has made prices go up as ISPs and datacenters now have to lease them instead of having them assigned (which only was a one-time fee in the past).
I would say that it depends on your location. Considering current average bandwidth cost, only European and North American based nodes would be able to turn a (slight) profit at $5/TB of egress (if big enough as small to mid size facilities rarely pay less than $10 per TB of bandwidth). But in reality ingress traffic also needs to be accounted for. As I mentioned in my original post, the current earnings calculation only really makes sense for individual SNOs with fixed internet rates. At $5, taking ingress bandwidth cost into account as well does not make anything sustainable. As a side note, Asian and Australian nodes are already largely unprofitable at $20 per TB egress (1 TB of bandwidth in Asia costs about $25, $35-$40 for Australia) thus the lack of nodes in the Asia-Pacific region on the network at the moment.
This is a very delicate topic as unfortunately, real bandwidth operation costs secrecy has kind of been made an industry standard over time. The only way to get real answers is either directly working within the industry or having insiders as you won’t find any good and up to date resources on this topic anywhere else.
My use is small and not representative. I used to pay until it got completely covered by the free tier. It’s inconsequential for the larger average. Which seems to be about 6% egress on static stored data. But synthetic data has had basically 0 egress for a while. This is not going to compensate for the payout drop, but it would compensate for a static storage drop if removed gradually over time. I see the payout drop as a separate issue. Only thing they can do is do it slow and carefully to ensure not too many nodes leave at the same time.
I appreciate you sharing your insights. I think this angle wasn’t sufficiently addressed in prior discussions. But it will eventually be an important part of remaining sustainable long term.
One thought here would be “safe” endpoints and “all” endpoints. If I’m using Storj for hosting of completely public data, I’d be fine with the possibility that one more person is accessing it.
Similarly, if I’m backing up data that is already encrypted, I don’t actually care if a third party has access to it, the double-encryption is not useful to me.
Why would I choose this option? Perhaps lower costs, or perhaps faster speeds could be achieved with a wider set of S3 servers.
I’m personally indifferent because I don’t use the S3 gateway anyway.
I’m still reading the entire thread, but I have a question that maybe someone can answer: what is the criteria for the lowest range of payouts?
I see in another thread, the one discussing the possible removal of the /24 subnet limitation, that one advantage of that is avoiding centralization, as Datacenters could easily allocate a lot of contiguous addresses.
But what do you think it will happen if payouts gets that much lower? Who’s going to afford having nodes running, if not Datacenters with huge scale? It looks like the proposed changes are not considering the decentralized aspect of this project.
I’m not against a payout reduction, as I stated elsewhere I’m here since V2, it’s a long journey and I believe the project has somewhat solid foundations. But if the decentralization gets weaker the whole sense of this project, at least for me, goes outside the window.
‘How’ much decentralisation is enough?
Currently there are 22k nodes. If we assume 2000 of them are not viable on the proposed payout changes would 20k nodes be enough for you to think of Storj as decentralised? How few nodes would you be willing to accept for Storj to still be considered decentralised?
As I stated in my message, it is not the number of nodes, it is the location of the nodes if all the propositions go live in production. We could have 200k nodes, but if they were distributed between 4 huge datacenters that would mean no actual decentralization.
It’s funny though how this confirms my point about resources you can find online about this topic:
The articles you linked all are 6 to 10 years old.
There are 2 key points here:
Many people (including the source of the graph you show) assume that using the internet becomes cheaper and cheaper because ISPs have been lowering their prices since the birth of the internet. These prices have been going down because the amount of people connected to the internet has grown exponentially. Another explanation is that the global network is getting more stable and mature thus reducing the maintenance costs. But remember that what ISPs charge is not at all representative of bandwidth cost for the reasons detailed in my first post.
The ipv4 runout (which is only getting worse over time) mainly affects datacenters and tier 1 providers as consumers don’t need public/static ips and it’s easier for ISPs to assign ipv6 addresses to consumers. A market with limited available spots (namely ipv4 ranges) creates competition. In order to win this competition you have to put more money than others on the table. As a result, prices are going up (NOT ISPs prices but bandwidth cost).
I mean, the graph doesn’t even show bandwidth, it shows speed. And while consumer bandwidth usage has gone up, it hasn’t gone up as much as the speed has. So technically people just use less of their available bandwidth. I also wonder what plans it’s based on. It didn’t cost me a whole lot to double my connection from 500mbit to 1gbit and that’s probably because I didn’t start using any more bandwidth at all as a result of that switch. So if they base it on the fastest connections, it’s really not that representative of bandwidth.
It’s also ‘funny’ because you don’t know that I sell business internet connections and bandwidth in the UK. So I am very aware of bandwidth pricing for the UK market. Whilst not everywhere in the world is the same I also have access to European and US pricing to know the long term trend is similar.
You mentioned the 2000’s so I found multiple articles which reference your timeframe. If you want something newer, well that’s available too:
I do agree with your comments about consumer ISP costs not being reflective of actual costs due to the market being about high volume (of customers) vs. low margin.
Again, I’m not seeing this at all. Yes, ipv4 address costs are increasing (i.e. they used to be ‘free’ but now can be sold and traded) but it’s not related to bandwidth in any way. You don’t buy 1TB of bandwidth with 1 x IPv4. They are totally independent, not interdependent.
bandwidth will keep getting cheaper so long as network technology keeps advancing at a rapid pace…
one day that might not be the case… but has been advancing for a long time…
ofc seen from a home LAN perspective the 1Gbit connection has remained for a decade or more and hasn’t really gone further… so from that aspect networking doesn’t seem to have advanced that much.
but it really has in the enterprise and ISP levels and is continuing to.
saying bandwidth prices has gone up is a misconception, tho the last few years corona and such, has caused a bit of a price spike in ISP costs and not helped by the inflation we are seeing now…
but its difficult to make an argument that it has really gotten more expensive…
and if so its a temporary little bump in the road… which seems to be eternally down hill
ISP’s had a lot of extra costs during corona, which they had to move on to the customers for the time after, doesn’t really mean bandwidth has gone up in price…
I never doubted the fact that you know what you’re talking about. My experience as a datacenter operator and bandwidth purchaser in Europe and NA is different from yours. As you have access to pricing data then you may agree with my original point made in my first post stating that the proposed egress pricing rate is not adequate considering current bandwidth cost in most parts of the world. I’m not even talking about infrastructure maintenance costs here, just bandwidth.
I didn’t say that they were correlated. Of course ipv4 now represent an entire separate market. My point there was to explain once more how individual and “professional” SNOs don’t have the same expenses as a whole not only talking about bandwidth. I agree that this particular sentence may have been misinterpreted and not clear enough.
I think it was one of the point of Storj that residential internet users have no bandwidth cost against datacenters. So it is static cost for operators not related with amount of bandwidth.
