According to stats storage_total_bytes vs storage_total_bytes_after_expansion is 1.4 atm. Select with much lower RS numbers is included.
What I wanted to say is, even with rented servers it makes no sense to use other RAID solutions.
RAIDZ2/RAID6 with 18 drives. Expansion factor 1.125. Or, you can go to RAIDZ3 for more fault tolerance, but 1.2 expansion factor.
You can adjust RS numbers to get the same or better with current implementation. BTW: ZFS is using RS encoding as well.
Yes, you can, but will the data be safe if it’s stored on the nodes? As I understand part of the expansion factor being bigger is that nodes randomly fail or get shut down if the operator gets bored.
If Storj decides to store data itself (no more nodes), then it can reduce the expansion factor by a lot and still be reliable. That was my point.
In today hardware prices it would be very hard.
In my opinion thats’ storj’s biggest asset atm… They have vendors ready to supply space at same or lower price while all competitors fighting skyrocketing hardware prices. But for some reason they don’t manage to use that opportunity. 
I don’t think S3 support needs to be removed, but for large clients, a warning could be issued stating that data transfer exceeding 5 TB (as example) per month via the shared S3 instance will incur a charge.
At the same time, a Docker image or Linux executable should be available, allowing users to spin up their own S3 instance on a local network with a single command and connect to it using any compatible client.
Instructions should be included with the notification so users don’t have to go looking for them.
good writing, any suggestion?
Instruction and docker images are available for a long time
Well basically we have 3 products:
- The object storage
- The hosted S3 gateway
- The self-hosted S3 gateway
Currently 1 and 2 are bundled and customers take them for granted. Storj pays for the hosted gateway and I am sure most customers take that for granted and don’t even know the other options. Also why should they? The price is the same. And why shold they add complexity and additional cost to their own side if they can have the same and better service for free from Storjs hosted gateways? So they won’t do it. It is very logical and rational.
To make them self host or integrate the native protocol there must be advantages for them. These need to be identified and adressed. It could be a monetary advantage or an advantage in form of transfer speed or whatever.
In any way the advantage must outpace the drawbacks that comes with it for the user, which is engeneering work, support, monitoring, maintenance, risk of errors, diversifaction, etc. etc. etc.
For Storj there is the risk that without S3 or a without competetive S3 product companies will not consider the Storj object storage as an option.
Wow, I am positively surprised that killing S3 does not stir up more controversy here. That is a very good change compared to a maybe a year ago.
Seems like some people are coming down from the VC money high.
Welcome back to reality 