1:how does safenet know the difference between a single large vault and a number of small vaults that in actuality are from a single server or a server farm?
2: Same Question, but now the big server farms are running their data through a VPN so that their vaults look like they’re from multiple places, and the server farms are colluding to this easier?
EDIT: And now I remember the other question: Why can’t safecoin be divided into 4.3billion parts at the start? It was mentioned somewhere that this would open up a method of attack, but I never saw an explanation of what type of attack would then be possible.
The system doesn’t care if there are multiple vaults on the same h/w.
When a different vault has to be picked up for any reason (such as replication), a random vault is selected.
The chance of that random vault being on the same physical system is very small, and since multiple copies are made, the chance of all replicas landing in vaults on the same system is extremely low.
The SAFE Network doesn’t know or care. A server farm running a bunch of vaults would not be bad for the network. Early on, it MIGHT be possible for a huge bunch of dedicated resources in a server farm to gain a large percentage of Vaults and cause disruption by shutting off all at once. But I doubt that will be done and if it is, it’s a test that the network will have to overcome on the way to more strength.
The amount of resources necessary to get some sort of control advantage would be prohibitive, I think, from rather early on. So all those vaults would have to behave or be disconnected. If they behave, they are contributing resources to the network and should be rewarded, as they’re just making it stronger.
Pretty much the same answer, except to say that the network also doesn’t know or care where the machines are located.
I think I should have pointed out that for questions one and two, I was wondering from the prospective of how the network implements it’s desire to prevent big server farms from getting the lionshare of the safecoin rewards for want of a better phrase, and not from a protection of the data issue.
Oh… kay, but then that kinda makes the entire topic of the profitability of big safenet farms moot, since big farms could just host hundreds of thousands of vaults per farm, and get big server farm amounts of safecoin rewards that way. You learn something new everyday.
Remember that a number of factors will affect the “profitability” of “server farms”
For the user who sets up a “home” server farm with 9$ computers then they will have limits due to bandwidth and costs
For the commercial farm it is questionable whether a farm can be profitable. Overheads are the killer, and ROI is too low for the equipment, when the equipment can earn more money (and more stable) doing other things, like being rented out.
The vault economics are designed for the home user who supplies excess resources which means little to no extra costs.
Allows slower vaults to get rewards (you don’t have to run out of data center because you don’t have to be one of fastest vaults to get GET requests)
Doesn’t severely punish downtime (eg you could go offline for hours without being kicked out and vault invalidated)
Vaults of above-average size get slightly fewer rewards (it seems based on the docs). You can have more smaller vaults, but that increases your overhead so it marginally still helps the small guy
I think the first measure is most important for this, to not always direct requests to the fastest vaults. Of course the same measure means the overall performance of the network will be slower, but that’s the price of “inclusiveness” - you always get the average or slightly sub-average performance.
Maybe I’m missing a technological limitation somewhere. However, if I’m not, and if maidsafe can’t tell how many vaults are coming from the same machine, and the network wants to only incentivize above average performing vaults instead of the highest performing vaults, then the server farms can just split their resources into many smaller vaults or combine their resources into fewer bigger vaults algorithmically based on want the network is demanding at the moment and get the optimal payout at scale.
Joe sixpack public isn’t going to buy a computer upgrade everyday and the network can’t make reputation ranking too hard to obtain in order to keep Joe sixpack’s resources worthwhile to farm with, assuming vaults are still non-persistent, so I’m not seeing how there’s a problem with commercial farming, unless the safecoin payout is keep to just barely more than the electrical costs.
The most obvious “problem” with it is competing with vaults which have no variable costs, because they are already paid for (equipment purchased, power & bandwidth paid for as part of normal use). If lots of users farm on their existing devices - which we can expect because it will be easy, and will help pay for their use of the network - it will reduce farming rewards making it harder for those who do have real world costs (pro farmers) to make a profit. Some will, but the effect should be to ensure that it is hard, and that everyday users will not be squeezed out (as happens with bitcoin).
No, because each chunk has multiple copies stored at different nodes, only the fastest (read: most valuable) node with that chunk will receive the farming request.
But also yes, because in certain cases the slowest node will earn a farming request from being geographically closer to the requester (network lag will be muuuuch lower = fastest chunk retrieval).
But I think the value appointed to each node by it’s close group is more important than internet speed alone.
The node value is calculated basically as how fast/reliable/useful is that node (at least down the road, for now I think it’s simply calculated along the lines of stored_chunks-lost_chunks).
If you store with SSD, RAM cache, and serve up over fiber, you’ll probably complete the GET request faster (faster file retrieval and upload), and I would argue they’d beat any lay-man spinning disk drive rig connected over wifi.
I don’t know what’s the source of your info, but in posts on this I’ve always seen that requests will not always be sent to the fastest replica holder.
And regarding geo-optimization, there is no way for the SAFE network to know geo location of any vault or node, “distance” is based on its internal address space.
I thought when requesting a chunk you’re asking for it from the highest valued chunk holder. Is that not so?
As for geo-location I did not mean the network knows where you are, I meant that if I have to move electrons 5 miles to the next town, that will be faster than moving electrons 5000 miles. The network doesn’t know location, but the length of wires (and intermediary servers) separating you and the requester still plays a role in the file distribution speed (and therefore plays a role in satisfying the GET request).
I actually envision - a decade or so down the road - corporations turning their datacenters into farms. In that case, they will be able to keep all of their data on the SAFE Network, and use the datacenters (as farms) to pay for that data.
One thing to remember (as I believe has been mentioned in this thread) is that the network forces the creation of many smaller vaults instead of having one huge vault.
There may be a script created that adjusts the volume of those vaults dynamically to ensure maximum payout-to-space ratio.
Where some may see that as gaming the system, I choose to view it as good ol’ fashioned intelligent system administration.
The biggest fear - vaults being close together in groups or having a majority in groups - is already addressed by the use of XOR space instead of physical space. If there are big farm datacenters, I can only see that as being both beneficial to the network, and beneficial to those who need to keep mass amounts of data on the SAFE Network.
The reward centralization will not exist however. The algorithm (that has yet to be implemented BTW) should (really: must) prevent any negative repercussions to Joe Sixpack solely because a big farming datacenter exists in the first place. AKA - it must take into account bandwidth also. This cannot be ignored.
Certainly different than what we’re currently seeing with Bitcoin.
It can be, if the parameter is flipped that way, which is up to the devs.
Because for quite a while David has been proposing that the network should not reward only the fastest machines, it seems logical that the algorithm would not be so simple to always fetch data from the best machine (which would very quickly lead to farmer concentration).
An “inclusive” algo would fetch the chunk from one of top 3 vaults (leaving the slowest one out), but if all four are ranked 20% below average or better, then to any of those four. This is just how I imagine it’d work.
I can’t provide a link because it’s all about the implementation - how the client or network picks where it fetches the chunk from. If that has been completed, then you could surely find it on Github, but I don’t know which repo and which part of code to look at. Maybe one of the devs knows.
I wonder what the value trade-off is to the public when SAFE purposely slows your downloading for a broader reward system.
You would think more people would jump on board to profit from farming, but at what cost.
Is the value loss from requesting chunks from slower nodes worth the value gain of a slightly more distributed farming system?
In reality the end-users will use whatever service does what they want the fastest, from downloading movies, to viewing large PDF manuals, to online gaming.
If it’s impossible to get below a 250ms ping in SAFE’s port of call of duty, then I would never use SAFE for that service, thus limiting the potential scope of the network.
My take on it is that bandwidth speed should be highly valued. Because of the rUDP protocol and distributed chunk allocation, the network should be blazing fast as it is, and if the network continues to reward the fastest nodes, it’s possible every single SAFE app/website would load dramatically quicker than it’s clearnet counter-part.
Leading to faster adoption among a wider range of industries.
Also something that has to be factored is that file retrieval is parallel not serial, so when I want the whole file I don’t need to request the first chunk then the second, but rather the client will requests all chunks and the file will only take the time to retrieve the slowest chunk PLUS/INCLUDING the actual speed of MY internet link (to transfer the chunks to my computer).
I hate to boil things down to comparing economic systems, but this is a case of capitalism working exactly as intended, it rewards whoever provides exceptional value.
If the service provides more value to me than other services, I’ll use it, otherwise I choose not to give it my time or money.
IF bandwidth was highly-valued, and nodes with above average speed were farming a majority of safecoin, then that fundamental principle of capitalism says people wanting to make money farming will buy faster internet (obvously based on the $/safecoin).
This in turn makes the network average faster, forcing people to switch from copper to fiber, making the average speed faster, etc…
The majority of people over the network’s lifespan will join to use apps. They will not be programmers, and they will not be interested in computing safecoin profitability on different hardware.
They’ll install it almost with spite simply because it’s delaying them from getting to their safe-netflix account, or their safe-facebook.
Speed is the most valuable resource on the network, because it’s so hard to come by. Everyone has 50+gb free on their harddrive, but only a lucky few have google 1gbps fiber.