It’s run it’s course. It’s just a free for all now I think, haha.
I suppose your right there, but there would still be far fewer nodes and fewer data connections contributing to the one real advantage Storj has to work with.
this just confirm, that most of people here to make money.
Also i think Storj want to here more different node operators, not the same just repeating same thing several times.
4 Likes
I think it’s wrong to assume these were necessarily the same people. Some people did compare Storj to others in the market, saying it was too expensive, but I definitely don’t have a memory good enough to know that those are the same people complaining now.
I for one did think Storj had to lower their prices, but was a little surprised to see by how much they were lowered when the new customer pricing was introduced. So there is also that in between option. That said, the customer price change is a done deal and I wouldn’t advocate for a blanket raise now as you want to offer customers some pricing stability.
I think this is just mentioned as a possible upside of these changes to node operators. Lowering the supply means nodes will fill up faster. But I don’t believe this matters all that much, since it will never be enough to compensate for the reduction of payouts and it doesn’t help full nodes at all. I honestly think this is a distraction more than anything else. Yes, it’s a small upside, but it’s minor compared to the immense downsides of this change.
I think one problem is that there are many much less efficient setups. And there is an argument to be made for pushing those out. In my mind, there are roughly 3 types of setups.
- Already on machines with existing HDD’s: Energy costs are negligible and limited to just the raise in power consumption between running a node on the hardware or not. This totals 5W for many setups. There is no expectation to cover HDD costs.
- Energy efficient dedicated storage: HDD’s spinning exclusively for Storj, hence needing to cover their own energy costs. Negligible or no energy costs for running the server, either because it was already online anyway or it is really low powered (rpi for example) or covered by solar panels etc. There should be a reasonable return on HDD purchases, but no ROI on server purchase costs is expected.
- Full server: Regular to high server energy costs + HDD’s dedicated to Storj. In some cases with SSD cache, high powered CPU’s, RAID and thus extra HDD’s slurping up power etc. Expecting a return on HDD as well as server purchases.
In my mind, it is reasonable to push out that last category as it’s kind of an overengineered for this purpose to begin with. But I think Storj should make sure the first 2 are profitable. I have nodes running with both type 1 and type 2 setups. So I’m not without bias.
It’s super trivial to create multiple email addresses though. You don’t even have to register new ones. You can just use user+storj1@gmail.com user+storj2@gmail.com etc. and they all end up in the same mailbox. There are other similar tricks as well and even actually registering separate email addresses isn’t hard. This wouldn’t be a good alternative to the /24 filter and it would again limit the number of legitimate differently located nodes an honest person could run.
Partially true. Common providers for cheap services will at some point also have nodes on most of their /24, limiting the impact this can have. But it seems at least for the moment it still allows some very big whales to get a large chunk of the data.
I have a feeling getting too many nodes won’t be a problem soon… quite the opposite.
But the situation for customers doesn’t change with this proposal. Assuming Storj is suddenly going to attract different types of customers without making customer facing changes seems foolish. No matter how hard marketing and sales tries, significant movement on this seems unlikely.
Well that just shows the impact of money incentives. If less efficient setups can make money, they will be created even if it was always against the advise of Storj Labs. But it’s reasonable to expect that to change with new payouts. Regardless how much it sucks for some node operators.
So do you consider lower oversupply as reasonable compensation for the drop in payouts in your scenario? It seems you have pretty much one of the best fits for these new payouts as long as you are able to fill up your nodes fast enough.
Easily saturated my connection. This is simply false. You could saturate connections 10-40x the speed of that with the correct settings and use case. Possibly even more since the parallelism practically scales linearly as long as you have enough data to transfer and the hardware and bandwidth to support it. See my previously posted tests on the “elephant in the room” topic.
I mean, I did have 10TB free… but was planning on using it at some point when I started.
I would not assume that. These tend to run on dedicated hardware and have additional costs for VPN/VPS. Furthermore, they use used HDD’s which have shorter life spans as well. Payout squeeze could make them no longer profitable, at least some of them.
New synthetic data hasn’t been added to the network for a long time now. And existing synthetic data has already started to be slowly removed and is no longer downloaded. These costs will be gone soon.
Long term, it simply isn’t. I consider this cost mostly gone as it will never scale up again, while actual customer data keeps growing. It’s good to get rid of it, but it’s a drop in the bucket over the long term.
2 Likes
i’m definately in your class 3 setup.
but i’m down to 2 watts pr TB at the wall, including router, networking gear and a few other things, which serves a double purpose or not even storj related.
and soon i will be down to 1.5w pr TB.
also if we calculate the power costs in relation to the storage costs, then for my case with 18TB HDD’s, it would be 8 watts pr disk x 192 hours x 30.4 days in an avg month for a total of
5836.8 watts.
so 5836.8 watts pr month x anywhere from 0.10$ to 0.5$ pr Kwh = 0.5$ to about 3$ in power costs.
a 18TB costing 300$, then it takes anywhere from 600 months to 100 months for an 18TB HDD to use the power cost equivalent to its purchase costs today.
this means that looking at the power costs is really a rather small fraction of the storage costs.
as each TB used will have to earn more than the HDD TB costs within a HDD life time.
i know these numbers drastically change with smaller disk sizes, but 18TB is the cheapest pr TB cost HDD’s last i checked, so they would be the competition.
i will say that my setup is getting into a bit of a CPU bottleneck 
so that doesn’t hurt my total watt consumption, even if they are rather inefficient compared to modern cores.
and it’s only really an issue when running the filewalker on all nodes.
but my point being… i think all this looking at the power usage is a mute point as the hardware and replacement gear is the highest cost factor.
sure lower wattage helps the profits, the total profits if low can be greatly benefitted by a 25% reduction in power consumption, because what is saved is pure profit.
I’m not here to give money to StorjLabs nor do i expect StorjLabs to pay me an unfair amount for my services.
and i think thats what you are almost saying, when you say one had the capacity anyways.
sure thats nice, but its still a resource you are potentially giving away for less than its cost in the current market assuming new proposed Storj payouts.
ofc back then storagenodes was well paid… so it was a non issue.
certainly would be much more difficult to justify starting a new node at the proposed prices.
but if one looks at the watt consumption and not the hardware costs, then its easy to think its earning… but really one is only getting a return towards the hardware purchase value.
again not saying you make this mistake, just that we seem very focused on the whole wattage consumption when it can be a very small part of the equation.
My statement was about the idea of reducing the prices for customers - at the time I think everyone understood that if the prices for customers go down, nodes will get paid less as wel, but there was hope of increasing traffic.
If you never planned on using the space, you probably would have bought smaller drives. Since it is not possible to shrink the node, any space that the node uses is pretty much lost. This is not so bad, if running the node pays enough to maybe buy a new hard drive if you need the space later. If, however, it pays $8/month, maybe not that good and instead is just an inconvenience.
Still, how many people have that much space they don’t really plan to use? Maybe a lot, I don’t know.
My setup is a some kind of mix between 1 and 3. I use the same hardware (including the same hard drives) for other stuff as well, but the node takes up the vast majority of space there, so if I did not have the node, I probably would have done things differently and would pay less for hardware and use less electricity compared to now.
I love that you explained about power, my setup is much more efficient but alot less actual storage so it couldn’t pay for new hard ware at the current prices…
3 rpi 3 different public ips 4TB x2 and one 10TB powe costs are close to nothing but they also only make 7 to 10 dollars a month so if one hard drive does die it doesn’t really make sense now to buy new hard drives and spin up a new node… And imagine at the purposed prices how much it wouldn’t make sense to buy new hardware for storj. As it stands right now it doesn’t pay for any kinda maintenance or to maintain any kinda downtime costs money that you won’t make in a month.
Imagine your a new user who’s got 500gig to spare for storj cause they love the project but he’s got a 4TB hard drive but decided to use it. Few months later jimmys hard drive with all his personal data dies because jimmy doesn’t know anything about raid setups to save his data. Jimmy then has to figure out how to get all his pictures of his family because he doesn’t have any space to back up anything (maybe he should of used storj) He proceeded to look for a company to get the data off his drive because the pictures he had were priceless but to restore the data would be 500 plus dollars.
Morally of the story is don’t use your unused space because that is ffffin stupid.
So realistically if you cannot maintain your node and replace hardware with the money you make there’s no storj and future looks very grim.
3 Likes
With S3? Saying that it is false is simply laughable. I just ran some test on my TrueNAS and got more than twice the speed for OneDrive than for STORJ. Yeah maybe you can saturate 1Gbit if you only use one file and no S3. That is not the point.
Fact is: that my dataset, with my 1GBit ISP, the native OneDrive TrueNAS integration was twice as fast as the STORJ integration. And the these OneDrive, Dropbox, Google Drive services are slow to begin with. But somehow STORJ manages to get beaten by OneDrive 
(Disclaimer: I only tested OneDrive none of the other cloud providers. No, I will not thinker around with tunables. Like most customers I simply care how fast it runs out of the box.)
We are talking about tool for developers. Do you expect command ffmpeg.exe cute_cats.mkv to do whatever you feel it should do without passing parameters?
And I may agree that defaults could be tweaked to be more “parallel”, but c’mon, the parameters are there for a reason. Find best combination for your use case.
3 Likes
Most clients are IT, that used to read manuals. Not granny that use thing out of the box only for 5% of its capabilities, because no one showed how to use it.
3 Likes
That is the problem is there’s not enough devs who want or need to use storj over every other company. Ease of use is a lot more important then the tinkering to get the best performance. The software should run a benchmark to figure out what the best performance would be not to spend hours trying to figure it out. Time is money.
1 Like
I am in a similar boat to you. Definitely class 3.
112TB for about 250W in a datacenter.
I pay about $110/month for colocation with 20TB bandwidth.
So when full $1/TB costs for storage.
but the egress / ingress cost is all lumped together at like $3/TB.
If the payments were cut so long as the storage was filled I would still probably make a small profit, but wouldn’t be making any return on hardware costs anytime soon.
How anyone would be incentivized to host a small node on a domestic internet connection with the proposed pricing is beyond me.
If there is not enough reward nobody will invest anything and you will stagnate the network growth.
The current conditions for running the nodes with their high IO usage and uptime requirements mean that you can’t use the drive for anything else, it has to be in some dedicated appliance that is always on. Slow upload speeds on domestic broadband can be saturated by the node. So you might want to turn it off for the time you use the internet each day, and it hardly makes anything anyway so who cares if it gets turned off, or you forget to check it’s still working, don’t replace a drive you really should… without an incentive it is not worth the time to maintain that hardware.
My point is the network isn’t going to get to the size you want/need without incentive to invest and maintain that investment. Whether that be in a better internet connection, dedicated NAS devices, etc.
2 Likes
may be it will be good idea, make some compiled version that will test your conection and will give you best setup parameters with one click
3 Likes
I think all options should be on the table to increase customers, I think the fact storj was only targetting Developers rather then everyone can use storj, I mean I use it for running some movies for plex because I know the possibility is there storj just needs to speed things up. But requiring edge services for speed is no good, I can clearly handing all the speed myself on my 3gig internet…So benchmarking it would be the best bet to use as much as it can.
4 Likes
So i will propose that storj make benchmarking tool working with 1 click for several platforma, to determinate best setups. Also i think it can be don not only by storj.
2 Likes
Hello Storj Labs
My concern with new prices ($1.5/TB) is that 18TB HDDs like the Seagate Exos will be interesting (due to low price per TB) while small common HDDs like my 4TB HDD Ironwolf for NAS becomes less atractive.
I don’t know if the Storj software runs well in 7200RPM 18TB drives. I will run the experiment in one of my exos just for having the detailed knowledge.
Edit: I haven’t done the numbers, but I don’t see any bottlenecks at first glance.
Edit2: Maybe moving the databases to a NVMe if the disk usage is too high.
This would make the storagenode even less profitable if you have to use Nvme drives…Cost per TB is alot higher then if you just bought rust. Even then Rust isnt that cheap.
You literally quoted me saying that I think Storj should support setup type 1 and 2. Where 2 comes with an expectation to cover energy and purchase costs for the HDD’s after a reasonable amount of time to start getting ROI. So no, I wasn’t saying that.
To be fair, there was some increase, but it wasn’t massive. But I see your point now, I misunderstood you before.
I agree, though it was my original intention to GE some satellites. But then I found out it had paid for an upgrade before I even needed to think about that. I’m afraid those times are over after this change. I likely wouldn’t have expanded to this size if the newly proposed payouts were in place at the time. But this is why I advocate for building a feature that automatically scales a node to fill up until 90% of the volume is in use and automatically scale down if other usage of the volume bumps it over 90%. That way Storj kind of disappears into the background. As long as it doesn’t grind your hardware to a halt.
I mean… how many people now buy 16+TB HDD’s? You probably have that kind of space when you first buy it. Or when you first build a RAID array (for other purposes) as those tend to be scaled for future growth. If a storagenode could use that free space more flexibly, I could easily see just choosing to make some extra money on that free space until you need it.
I’d say the moral of THAT story is make backups of your personal data, haha. I’ve argued against RAID before as you well know, but I’ve always made an exception for the “use existing hardware” use case. I run my internal nodes on RAID6 (technically SHR2) myself, because it’s using extra space on a multipurpose array. But I also have offsite copies of all personal data on there. That’s just basic backup strategy. And yes, one of those copies is on Storj. Sorry for the node operators, but I never download that data. No egress from me.
Ahh, the hallmark of someone proven wrong… adding qualifiers after the fact. You never said S3 originally. You made a blanket statement that Storj isn’t that fast. Of course I advocate for using the much more scalable native integration. But if you insist on S3, you can still do it by hosting the gateway ST locally and tuning both the gateway and your S3 commands to use high parallelism. I’ve not tried it with gateway MT, but you can use parallelism with the S3 connection there as well, but due to the nature of the added bottleneck, I doubt you can have that unlimited scaling doing that.
You can saturate a whole lot more and you don’t need to have a single file either, as you can parallelize across files as well.
You can put this statement between brackets all you want. But this tells the whole story. Storj is aimed at developers, not end consumers like OneDrive is. And yes, it requires some tuning for optimal performance, which developers are much more likely to be willing to do. Developers aren’t like most customers…
I do agree with that. And when I previously mentioned the uplink should default to higher parallelism, this was exactly the suggestion a Storjling made (I forgot who). So I guess they agree as well. But I fear with 95% using the gateway, this hasn’t gotten priority.
Seems that line speaks for itself.
Yet all my biggest nodes run on shared hardware, including shared drive space.
The easiest way would be to just start transferring with parallelism at 8 and add threads until it stops increasing the transfer speed or starts bottlenecking the CPU. Use the transfer itself as the benchmark. You’d also have to take the amount and size of files in mind though.
Have you considered having Plex use a local S3 gateway ST and tuning that for your use? Even though there is no pricing incentive to do that currently.
Would you try it if there was a pricing incentive?
Not if you already had those for other purposes. Also db’s do not take up a lot of space.
I have played around with it but I wanted to make it as simple as possible so I just decided to use a program called mountainduck that will mount the S3 drives in windows and just point plex to that hard drive.
But again not everyone just has nvme drives laying around in there systems we cant assume that everyone has a Powerful Nas that can support nvme drives right?
Databases dont take alot of space but still can put wear on nvme drives, Which I have killed 2 nvme drives in recent times because there using cheap chips that wear alot faster. So your luck could vary. So price per TB would still be higher if running it kills your nvme drives or nvme drive just fails.
We can always use the excuse as “you used it for something else” But happen to run storj on it but in reality that just tells yourself to make it worth while and fills your heart with joy.
Would you go back to that if there was a price difference?
No, I don’t expect that. I don’t think it’s necessary in most setups either. But use it if you have it I say. The wear for only the databases should be fairly limited. I personally run SSD’s as a full cache solution, that’s definitely not a reasonable expense for just Storj (which it isn’t exclusively for in my case) but yeah, wear is more of an issue then. I’m not sure if that’s a wise decision of me… we’ll see. I can afford to try it.