How to ensure data durability?

If only one miner stores my file, it is easy to be lost.
How can I store data with error-correcting code as described in white paper?

Redundancy and encryption are the responsibilities of the client. That functionality has yet to be implemented, while we’re focussing on the core protocol.

How N+M mode which described in white paper works?
If client works with N+M miners, it is low efficiency, and brings new issue: who takes responsibility of maintaining (such as rebuilding lost slice)?
If N+M slices are all located in same miner, how to avoid disaster which destroys the whole site? DR is one of most significant feature of decentralized storage, IPFS/FileCoin should not lose it.

You’re right about the importance. We intend for this kind of functionality to be implemented on top of the core Filecoin protocols (which are not yet complete). We expect redundancy and maintenance to be managed by some “maintenance miner” functionality that monitors and executes calls on behalf of a user. In the long run, such a miner will likely be smart-contract code (so there can be many implementations), but right now we’re focussed on lower-level protocol implementation.

Actually I have d same concern about it.
Lets say I pay certain amount of FIL to store some really important data on IPFS network (no backup), but my data goes into only one miner’s storage, if this particular miner stop mining due to several reason, even dont mind lose pledge, just simply disconnect for goods.
then I might lose my data for goods as well !
I’v read the issue listed by NeoGe “Do I need work with millions miners?”, he mentioned A theory “reputation”, I m interested on it, my understanding is that “reputation” can tell us which miner more steady than others, then we can chose miner base on it.
Is there any method to identify a miner with greater reputation? or any method to determine the level of miners base on stabillization and durability?

1 Like

Is there any possibility to open a web to show the ranking of miner base on reputation? then I will be more confident to store my data :slight_smile:

Yes. Because storage over time is proved on the Filecoin blockchain, that kind of reputation data will be relatively easy to gather. We haven’t implemented anything for it yet, though.

1 Like

Just wanted to say that the reputation of nodes can be tied to its uptime. So for the durability of a miner the network can simply use the uptime of the node. As the node has more uptime the node will be less likely to disconnect (referencing from the chart on page 5 of maymounkov-kademlia paper). Since Kademlia DHT works pretty well in real world applications, I think this reputation scheme will probably work for this senario. But of course uptime does not guarantee storage over time so maybe the reputation could be a mix and match of many parameters.

1 Like

What is the goal of data durability of IPFS/FileCoin? IMHO, data durabulity based on reputation can not reach same level of centralized storage such as AWS/Ceph/HDFS

1 Like