Can someone help me with this question… If for example my 4TB SSD has 800TBW (“Terabytes Written” i.e. the metric used to indicate the total amount of data that can be written to a Solid State Drive (SSD) over its lifetime, for anyone who doesnt know what i’m talking about…)
and I contribute this to the network, how quickly might this 800TBW get used up on the network - given all of this moving and shuffling that the network does. How does the network make sure that this total memory isnt used up so quickly that the SSD’s lifespan is shortened prematurely and therefore add additional costs to resource providers? Knowing this would help me as a potential resource contributor figure out what my total cost to contribute would be. and therefore what level of reward i would accept from the network.
I doubt that will be a significant cost element. More significant imo will be bandwidth and then energy use, and then maybe overall hardware cost but I think the first two will dominate. So I’d focus on those for now.
I don’t have exact numbers, but it should be completely fine, my guess would be 10-100x below level that can cause problems. If you want to be extra careful, you can limit logging or have one more durable disk extra for logs.
Personally I have seen dead SSDs from too much data written only in two scenarios - heavily used database and home PC with not enough RAM and permanent swapping.
Thanks @peca & @happybeing for responding. This one is off my list of concerns then!
I’d expect that nodes won’t reset too often at all once everything is worked out and the network is large enough. So it would be reasonable to assume that the drive will last a long long time and it’d be best to purchase cheaper drives with lower TBW capacities to maximize your return on investment.
Join one of the testnets and have an app (or OS) record the number of writes to the disk over say one day and then you will have a metric to base some real figures on.
Generally on top of the TBW figure is the amount of free space on the SSD. Each block can only be written so many times, used to be 10,000 and some now are 50,000 and this is the limiting factor more so than the generalised TBW figure which is an approximation based on the number of times you can write to a individual block on the chip
So if you have only 20% free then the cycling of blocks only occurs on the free space blocks of your SSD. A SSD when writing new data chooses the least used free block to write the data to. As you can imagine the more free space you have the more blocks that the SSD can cycle through thus extending its life. Many SSDs write the total blocks written to its S.M.A.R.T. registers and also usually have a figure for used life expressed in %age of life left.
tldr
keep as much free space as possible on an SSD for longer life of the SSD. If a SSD of double the size of your needs is like 20% more in cost then always get the larger drive and have up 10 times the life of the drive.
Yes I would expect this to be true. A node is writing new data when
- new chunk is written to the node
- Node leaves the network and its chunks are spread across multiple nodes.
- your node only gets chunks that that are now closer to to your node. ie your node is now in the 5 closest nodes.
- I would expect this might be 1/5 of the leaving node’s chunks
Then chunks are removed when a new node appears and the reverse happens of node leaving where your node is no longer in the 5 closest nodes for some chunks.
So while constant churning is occurring as nodes join and leave which are your neighbours, it shouldn’t be too much when there are 100K or millions of nodes.
Also using SSD as SWAP is bad
Better is buy more RAM and not to use SWAP at all. If I had to use SWAP for some reason I would put it on SSD without hesitation. Life is short to wait for a slow disk and SSDs are cheap.
I recommend not assuming that you can use all your 100Mb internet connection for Safenodes or thinking it’s a good idea even if you can. If you ever want to do anything else in the house that can require a lot of bandwidth like play games, watch Netflix, etc. you’ll find that one or the other suffers. Also, if the Safenodes become throttled by other uses or a general slowdown on the connection to the ISP they will not respond quickly enough and may be kicked out of the network. I’ll be aiming for 50-70% of bandwidth utilisation for Safenodes and no more.
If somebody create on safe network social network, will be there then personalised ad for each user?
I think the whole idea is that would only be possible if the social system had decided they wanted to do that and the viewer had consented to it and the use of their personal data and said what they want to allow to be used and what shouldn’t be.
What there wouldn’t be is snooping on location of the user, details of what files they hold on their machine, what other apps they use, what the data in those apps are, etc. Because that wouldn’t be possible. No facility to do that.
There is a place for personalised ads but you want to be able to control what data is used and how.
No need for ads as the developers will be paid in tokens.
how for example on social network will developer get paid in tokens? I dont understand, it isnt farming (with node)
From Discord:
JimCollinson — Today at 20:13
There are teams already developing social networks that will operate on the safe network. But they will use significantly different business models than the highly algorithmically surveillance-based ad models that support the likes of facebook. And that’s by design. The network can’t know, nor learn about you from your browsing habits. It’s built on privacy by design principles. Could you do advertising on the Network? Sure. But it would be back to traditional billboard-style, or product placement advertising, rather than highly targeted attention economy, click-bait inducing, behavioural modification models that we experience at the moment. And I think that will be a positive step toward a better Internet.
Royalties
I have a similar concern about the basic principles for long term storage. From the RFC 0061.
Users wishing to store data on the Network, or edit existing data, pay the Network to do so in Safe Network Tokens. A Data Payment is made upfront and there are no ongoing costs to maintain data on the Network after this payment—content is made perpetually available after this one-time fee. rfcs/text/0061-safe-network-token-distribution/0061-safe-network-token-distribution.md at JimCollinson-rfcs · maidsafe/rfcs · GitHub
My concern is as follows. Lots of people upload private data and this will be perpetually stored for a one-time fee. When this person dies this data stays online but the one person that could access it is gone, therefore the system will continue to store inaccessible data perpetually.
I can’t imagine this principle to hold over time, either financially, or otherwise ecologically. Is there a build in system to combat this effect?
Every system has waste. Growth of data works to shrink this effect.
Think about how much data you, we, everyone handled 5, 10, 15… years ago. It’s not long ago people had none. In twenty years how significant will all your private data be compared to what you have today?
Another effect is that we won’t need to hold as much data in private silos as we do now because it will be accessible as public data.
But to answer, is there a way to delete dormant private data… not at the moment.
It would be difficult I imagine, but if a way to decide could be devised then it could be implemented later. But I doubt it will be an issue for the reasons above.
I bought a Compaq PC in 1997. It had a 2GB hard drive. The staff in the store were all standing around in awe that someone had bought the top of the range model with crazy amounts of storage.
My Firefox app is currently using over 2GB of RAM.
This is the contract. The fee, which is not fixed, will be an estimate of the cost of performing the service. It won’t be exact though, so over time it will adjust to fit the expectation plus or minus past inaccuracies.
I’m aware of the advances made in storage capacity and data creation. But I’m not yet convinced that his won’t be an issue long term.
Just a quick and dirty calculation:
Deaths a year 80 million until 2100.
Personal data stored on average for a whole life: 20TB
So in the year 2100 there will be (2100-2024) 6.080.000.000 million people who died. With an average of 20TB makes 121.600.000.000TB of inaccessible data stored perpetually.
In 2100 the expectation is that there will be around 10 billion people, lets assume they have an average of as well 20TB of data stored that makes the whole capacity required for the next internet 200.000.000.000TB + 121.600.000.000TB = 321.600.000.000TB. So roughly 1/3 is wasted space, energy, resources and this 1/3 share will continue to increase because the total world population will decrease while the total amount of personal data that is inaccessible will only increase.
One might argue as @Astroman does that 20TB now might mean nothing in the future. But this comparison breaks because they both develop at the same rate only with a slight deviation.
And at the same time I can see a reduction in data growth happening because photo’s, videos and other sources of data don’t need to increase as drastically anymore as they did in the past due to our biological limitations. A 3D 16K 120fps video stream will likely be sufficient even in 2200 to match the quality of our eyes. And when our lives are stored continuously there is no additional “time” created so 24/7 is another limitation.
The numbers above might change drastically, but there is still, in my opinion, a very likely scenario that the amount of inaccessible data is substantial for the performance of this new internet. And if this network needs to support this burden a competitor who solves this can easily be 1/3th cheaper in running cost.