Then you can just buy HDD with smaller capacity to your library and save money.
You would think so. Like if businesses can’t afford their own space now… and hope prices will decrease in a couple years… would they “rent” more S3 space until then?
HDD companies say 2027 and 2028 production is sold already… 
Yes, because for company pay 50-100$ a month is much easier than buy 5-7k server+UPS now.
How repair mechanism works? Dos it audit pieces, if it absent for some time it marked as absent and if this pieces absent enough it go to repair que? It usually takes some time, but before repair, does it check maybe there is enough piece’s again, as offline was temperarly?
You want the extended answer. Sure.
For the record current RS numbers are 29/46/54/75 Not sure for how long. I feel like we changed the RS config around so many times that every time I write it down it is a different config.
So lets say a segment gets uploaded with 54 pieces. 4 nodes call GE, 3 nodes are getting disqualified and that gets us down to 47 pieces remaining. Now we take 2 more nodes offline. That gets us down to 45 and the repair checker will queue it up for repair. Lets put 2 of these segments on the repair queue.
Next the 2 offline nodes come back online. Both segments are still on the repair queue. Repair worker picks up the first task from the queue, checks the status of all 54 nodes, finds out about GE, disqualification and all that. It will notice that the offline nodes are back online and the segment is back to 47 pieces. → Skip it and drop it from the queue. Meantime the repair checker finishes a second run and will update the segment healt of all segments it has observed. In this case the other segment will be missing. That segment is still on the queue but the repair checker has not seen it and not updated it. → Delete everything from the repair queue that has a timestamp from a previous run.
In both cases the segment doesn’t get downloaded. Lets continue with our journey. This time 2 nodes have a bit flip. Not enough to get disqualified. The repair checker doesn’t know that and will not queue it up. 2 nodes go offline again. 45 pieces remaining and it gets queue up. This time the nodes don’t come back online and the repair worker has to do some work. It will download 29 pieces and might encounter one of the bad pieces. The repair worker also requests the signed piece hash from each storagenode. That way it can identify the bad piece, download a piece from another node, recostruct the file, regenerate all the bad pieces including the bit flip it has detected and upload them to new nodes. Write everything back into the database.
Any followup questions?
I am missing the word audit. What happens when I see a GET_AUDIT in my logs?
And how long it take depends from system discovered and repair process itself?
I thing it depends from repair queue length.
That would have to be “per satellite”, because there’s supposed to be no coördination between satellites on node use. So, first 500 GB on us1, first 500 GB on eu1, first 500 GB on ap1, and first 500 GB on saltlake?
You could just shave around 75-cents from each node payout and end up in the same place?
Not saying either idea is a good one: just a change that’s just an Excel formula is way easier.
Maybe. Not liking it myself for other reasons I’ve posted, but any proposal in this area must be very clear for discussion, because it affects the core network’s stability.
Why? Because a few users on the forum say so? I have drives that I bought specifically for Storj that have fully paid for themselves and are still generating profit.
By the way… what about this month’s payouts? 
Because here storj itself says: Storage Node Operators contribute unused hard drive capacity to the network
Because here storj says: you enable your unused hard drive space to be accessible by the Storj
Recomendation here) says: We do not recommend purchasing anything specifically for use with Storj alone
And so on and so on.
I still don’t understand the point. We all know Storj’s slogan and the vision behind the project. That should be its public face, and it should remain that way: a beautiful and ethically compelling concept.
But then there’s the real world. If I can earn €150 a year from an 8 TB node, and I can buy an 8 TB drive for €100, why shouldn’t I consider buying hard drives specifically for this project?
Anyway, I didn’t mean to open a discussion about that. I’ll keep reading advice telling people not to buy drives specifically for Storj, and I’ll keep not understanding the reasoning behind it. I’m fine with that. Probably you all have better business in your hands with better roi numbers
Besides, there’s no point discussing it now, when the whole thing seems to be winding down.
We have a minimum payout threshold, but it’s a different story and purpose - the payout should be great enough to be sent over blockchain.
The idea behind Storj is to utilize the spare capacity of existing equipment, not to build something specifically for Storj, otherwise it wouldn’t be any different from a data center.
But we also usually warn people not to buy equipment just for Storj; it might never pay for itself. Otherwise, the user will come back and complain, “Your payments don’t cover my bills and don’t allow me to get an ROI! I invested for you!”
And by the way, this is exactly what’s happening here on the forum. A user decided to invest in equipment, but demand shifted, the pattern changed, the number of nodes increased, and then the payout drops. But they already invested in the equipment, and now it’s burning up their electricity and their nerves, resulting in losses. Of course, they come back and complain that Storj is to blame, right?
No, we warned them that this could happen, so it’s less risky to run nodes not on specially designed equipment for use exclusively with Storj, but on equipment that’s already online but has spare capacity.
Or the equipment should be cheap enough that it doesn’t create a hole in the budget (interest and experimentation, i.e., a hobby, play a more important role here).
We are blaming Storj for not being able to sell the product successfully.
We wouldn’t have that kind of discussions if we had Exabytes of data on the network.
It was already the testing with 1 single customer (Vivint) that brought noticable stress on many SNOs.
And of course we blame Storj that they claimed how successful everything is when in fact it was not.
Oh yes, and Vivint signed for the Select network at the end. Not because the Global network did not meet their expectations but because of the lack of SOC2 certificaton for the Global network.
Yes, Storj is to blame for the current situation partly.
Then I would like to suggest you to lead the Storj’s Sales team to teach them. I shared all your ideas.
However, my answer was not about sales, but about a reason why we do not promote to build mining farms.
I see that. But you cannot separate the one thing from the other.
If you are reading threads from the past, then it becomes clear that Storj was quite happy that the back then expected demand resulted in SNOs preparing for it:
So I’d say anyone who did not notice that we are not talking about unsused ressources must have been living under a rock. Ressources were specifically purchased and added for Storj in expectation of the expected growth that was announced multiple times:
So at the end as always the failure comes back to the data that did not hit the global network not on the operators who prepared for it.
And also I believe back then the term potato node surfaced due to the many many nodes that could not keep up with the test load or complaining that their nodes are getting full. The assumption is that these mostly came from the “use what you have” SNOs. Because probably not many have 600TB of unused disk capacity lying around.
Hype precalculated and pre-arranged by management having a pre-paved exit strategy? Could that be an insinuation? Nooo…, definitely not. I concur. Yes. Maybe so. Nahhh. If you could only see what’s coming through your CCTV camera - don’t let the door hit you on the way out!
2 cents,
Julio