You consider adding outsourced GPU-computing. Isn't this a centralization risk? Is the access to the network impacted during outages of these central providers?
It shouldn't be a significant centralization risk. The costs of buying your own GPUs or starting a new outsourced proving service will be pretty low. Even if many people decide to use third-party provers, there should be a competitive and well-distributed market.
Recommended hardware is a huge entry barrier for miners. Is the design decision worth considering you'll leave out home miners with small home servers and NAS?
We're working really hard to reduce sealing and proving costs. We really want to support smaller miners, but not at the cost of security. It's always possible to make things faster and cheaper through later optimizations, but it's hard to make things more secure.
How long is the storage task lasting, if a miner use "lotus-storage-miner pledge-sector" command? how to avoid pledge losting in production chain?
Why here! We will make this customizable, right now in lotus its set to some pretty big value. We might even make a special mode for 'filler storage', storage youre adding just for power that doesnt have any real deals associated with it.
If the entire city has been out of power for 24 hours, how do you punish storage miners?
Details around faults and recovery can be found in the spec (https://github.com/filecoin-project/specs/). The protocol will accommodate both realistic operational needs for good miners and service providers, while maintaining the security of the protocol and Quality of Service for Clients' data. Having a data storage service offline is not good for Clients, and are not good for the service. At the same time, temporary power failures happen, and the protocol needs to accommodate them to some extent. As we get closer to mainnet, we will set cryptoecon parameters and policies, and we will explain them clearly for everyone.
Any recommendations on how miners should handle multiple disks in their miners? RAID, ZFS, etc?
Hey! Why here! I've thought about this a bit, and it depends on the amount of storage collateral you will put up for the data you are storing, if its a high amount, I would go with a RAID setup (personally prefer RAID10 as repair times on RAID5/6 kinda suck). But if you are a 'cheap' storage provider with low prices and low collateral, it may make sense just to store everything in a JBOD and eat the cost of the failures. Whatever you do, don't use RAID0 as a single disk failure pretty much loses all of your storage.
If I save a file in a miner, which corresponding sector isn't filled yet. How this affects retrieving my file?
Users can save a file before a miner seals it into a sector thanks to having deals on chain. The details can be found in The Filecoin Spec (https://github.com/filecoin-project/specs). The miner can then seal the sector when it is filled with deals and users can retrieve the file from the sector later. Note that users can also retrieve their file immediately before it is sealed into a sector. This is not currently implemented but we plan on doing this soon.
“outsource SNARK computation to third party service providers.”How to avoid the centralization of these third-party organizations？
- make it very easy to provide the service -- so that if anybody is being malicious or charging too much, others can provide it.
- Encourage many (30+) different service providers to exist
- you can always run it yourself -- if it is more expensive to use a service than do it yourself, you will do it yourself. -- this means that service providers have to charge fairly, or nobody will use them.
If a miner finds a mining algorism strategy, which can earn much more reward than the contribution of resources on the testnet/ mainnet, will that be feasible
The security of the protocol is Filecoin’s number one priority. Our protocol design and proofs teams have invested heavily to ensure the best way to earn mining rewards is to provide lots of proven hard drive storage. We’ll continue that work to make sure miners are always incentivized to provide useful services to the network. If you discover a great mining strategy, we'd love to hear about it on Slack!
During this test network, one company has exceeded 51% power. In this case, whether other miners have the ability to participate in the later stage of mining？
Hey, Why here! It looks like one company prepared for the testnet launch much more than anyone else. They are not misbehaving with this advantage, so new miners should be able to join in and gain power just fine. If they end up doing something bad (which i really doubt they will) we may have to reset the network.