node aging is progressing fast, feels like the finish line (network) is really close now… lets see what todays update will bring
I think they use a single IPv4 address to determine a “machine”(aka server) so you would need a “/28” IP block to pull this off.
Are there any advantages for the network letting multiple vaults run on a single machine?
I would say this was the opposite of decentralised, a single sever running multiple vaults = a single point of failure, the underlying physical machine.
If the farming rewards are calculated in such a way that running multiple virtual machines on the same physical hardware = more rewards (incentivising virtualisation), I personally feel this is damaging to the network.
The same underlying physical hardware should generate the same farming rewards and a server cut up into 50 smaller parts (In my opinion). Until we get to the ability to try it out we won’t fully know how physical vs virtualisation compares for rewards.
Just because a machine can run multiple vaults don’t mean that the network as a whole is centralized. If one machine with many vaults hold all chunks for a file or a majority of files, that would be centralized. But that is not what is going to happen?
The question about farming rewards might be relevant, the comment was just about general logic, centralized vs decentralized.
It looks that if you have 100TB and 1Gbps without multiple vaults you would not be able to fill that space in near future.
If you have small space to share there will be no advantage to run more than one vault.
If virtualisation is incentivised, then the problem is that everyone will do it (because greed). which inherently makes the network weaker not stronger. Encouraging any behaviour that makes the network weaker, is a terrible terrible idea.
My question is does multiple vaults help or hurt the network. This feature should only be added if it helps.
I do not think widespread virtualisation will cause any major issues. In fact more people will be encouraged to run vaults.
The person with a lot more $$$ invested in the current system will be more encouraged to run any vaults since they can run 2 or 3 or 5 or ? vaults and make better ROI. Just one vault that can never be filled will not justify the effort in many eyes of people with large systems
Its not like virtualisation will allow the home user to run 100’s of vaults except in the rare case
The larger the network the less the problem. For a 100 node network its a large problem, for a 10000 node network the problem is noticeable but much smaller, for a 100,000 node network then doubt it is quantifiable or much different to the whale that runs 1000 nodes in their datacentre
At what point do you declare the network is geographically diversified. What about a university dorm where 5000 students are all running one vault each? The internet connect is still a couple of links in/out. The power is fed by one substation. The dorms are affected by the same problems
This is a poor argument, one vault that can never be filled because there isn’t enough demand/data on the network is absolutely fine.
Tricking the network into think you are running 6 vaults, which are all on unique hardware, which are in fact a virtualisation, in an attempt to store more data and gain more rewards, is a path to destroying the network. You are increasing the chances of storing more than 1 copy of the same chuck, for your own reward.
A fundamental problem isn’t made better by 100,000 physical machines running 1 million vaults. It’s still a weakness.
Did you think about it first, that is not the reason.
Its the shear size of the vault.
And you will earn only one reward if you serve up the chunk quick enough. Thus its a disadvantage to the vault owner. Otherwise there are still 6 other copies
I did think about it. The network won’t store all 6 copies of the data on one gigantic vault. Thats how it’s designed to work.
Cutting that machine into 6 pieces mean you could end up with 2 copies of the data in that same gigantic vault. This is a weakness. Which means you are more likely to get the reward for serving the data because you only have 4 other competitors not 5.
I see the point you are trying to make but it does not seem to make fully correct logical sense to me. If you can make arguments in a deeper more detailed level that might help. But in general, if you have 1000 participants with 1000 or 100 000 vaults together, that won’t make a difference, it will be the same as the files will be evenly spread across the participants.
You switched responses. I was giving you the reason for why one vault will not fill - the link speed is the limiting factor. And that was indicated by the portion I quoted.
How can we discuss if you do this.
And my previous post pointed to the problem is one of network size rather then specifically of multiple vaults at one location. And where do you draw the line preventing multiple vaults. Do you say to the dorms they can only have one per building??? Or an apartment building with one internet link to the outside world, one power supply?
If everyone virtualised the same amount, the weakness is reduced, I can see that.
The problem is human greed, if virtualisation is incentivised, you are creating an “arms” race, in that farmers will be constantly trying to maximise reward at the expense of network strength.
It’s not even a difficult problem to solve. Once we get into beta I think what we find will happen is this. A vault will allocate 2TB of data to the network, and that space will be filled over-time along will node-ageing, at a consistent rate proportionate to the size offered to the network (so cutting it into 10, 200GB volumes won’t fill it any faster). This will either consist of “real data” & “discardable data”. The farmer won’t know the difference.
This means that you can’t over provision the same underlying hardware as a way to gain more rewards. Its not 100% fool-proof, but it levels the playing field between virtualisation & physical hardware. Reducing the impact of virtualisation is only of benefit to the network.
I would have thought that virtualisation will just be a mechanism to get good utilisation out of available hardware, which is pretty much its purpose.
If you spin up too many VMs/containers/vaults, they are going to impact on one another. As they fill up, some will need to be sacrificed, thus losing age and earnings.
I suspect the right number of VMs/containers/vaults will correlate with the continued capacity of the host. The vaults will then gain age, without being rejected by the network as they slow down.
Tbh, I would expect launchers to be created that start the optimal number of vaults for the host. This would create a level playing field and make best use if hardware in the long run.
The problem is that if your PC crashes then 16 vaults will crash simultaneously. This isn’t good for the network.
But is it not the same as having 20 machines on a LAN and losing the connection or power outage?
In the cases you mention it is the same, but not in the case of a machine crash. In this case you will lose only one vault.
But then too the university dorms where they have one internet link, one power supply to buildings.
Where do you draw the line?