Professional technical analysis of Filecoin (IPFS) mining

With the continuous release of Filecoin information from Protocol Labs, a lot of Filecoin mining information has surfaced. Filecoin mining is very meaningful for research. In order to help investors deeply understand the knowledge of Filecoin mining, Mingge from Star Continent has compiled some knowledge for your reference.

Ming Ge will analyze some elements and knowledge points of mining from the perspective of popularizing knowledge, and provide reference for maximizing miners' profits.

1. The Principle of Proof of Replication

Proof of Replication ( PoRep ) is a new type of storage proof scheme that allows storage miners to convince users and other miners that data has been copied to their mining machines. This scheme can effectively prevent storage cheating problems such as witch attacks, outsourcing attacks, and generation attacks. The following are simplified steps for proof of replication:

1. Miners create a specific replica R

When a miner stores a sector (this sector is not a physical sector on the disk, but a logical sector defined by the Filecoin network), the data in the sector is added with the miner's secret key pair, and a slow sequential encoding is performed to form a specific storage copy of the miner, making each data copy different. The result of sealing is a data copy with a Merkle tree structure, as shown in the figure below.

Data copy diagram of Merkle tree structure

2. Miners generate proof of storage data copy Πpos

After the miner stores the data and completes the seal, how can the miner prove to a Filecoin network without a centralized server that the miner has indeed stored this copy? Mingge from Star Continent found it difficult, but Protocol Labs did it.

The idea of ​​zero-knowledge proof is used in the process of generating the proof (zero-knowledge proof is a bit complicated, interested readers can read relevant materials), that is, the verifier believes that the miner does store a copy of the data without having to own any data. How is this done?

As shown by the red line in the figure below, based on the random challenge C proposed by other verification nodes in the network , for example, if the miner wants to challenge position 8 , the miner needs to calculate the path from leaf node D8 to the root node Root , output a proof Πpos , and return it to the verification node that initiated the challenge.

Miners generate proof of storage data copies Πpos graph

3. The verification node verifies the storage miner’s proof

After the verification node obtains the proof from the storage miner, it can verify it with the given information. When the verification result is equal to 1 , it means that the miner has indeed stored a specific copy. The verification process is very fast. The specific verification algorithm needs to be explained in detail after the official code is released. The existing white paper does not provide a sufficiently detailed description.

2. Explanation of sectors

Sector ( sector in the white paper ) refers to the storage space specific to the storage miner, which is used to measure the effective storage and the unit space for submitting storage proof. When the storage miner has stored a sector of data fully (it may be sealed even if it is not fully stored, but it should waste some storage space), the sector sealing will be run, and the replication proof will be submitted after sealing, so that effective storage can be formed.

Currently, Protocol Labs may have two solutions for sectors. One is a fixed size, such as 1G ; the second is a size specified by the miner, such as 1G or 10G .

So what is the use of the size of the unit sector? According to the description of the Filecoin blockchain data structure in the white paper, the sector ID will be stored in the allocation table of the blockchain, and one sector will consume about 300 bytes of on-chain storage space (this data is obtained by Star Continent in communication with the official, and is not displayed in public information). Therefore, it can be calculated that if the sector size is 1GB , after a period of time, the effective storage of the entire network reaches 10EB ( 10240PB ), then the storage space required by the sector on the blockchain is about 3TB . This does not include the blockchain space occupied by orders and transactions, that is, maintaining a Filecoin full node requires more than 3TB of storage space.

Therefore, if the sector is too small, it will take up a lot of block space. However, if the sector is too large, it is not easy to seal, which will lead to the waste of some effective storage. I believe that Protocol Labs will make a very good decision on how to balance the sector size and blockchain space.

3. Proof of Space and Time

The proof of space and time can be understood as a continuous proof of replication, that is, miners must continuously generate proofs (here, Mingge boldly predicts that it will take about 10 minutes) and submit storage proofs within a submission cycle (boldly predict that it will be submitted once every 3-5 days. The purpose of this is that even a short-term network outage will not affect the generation and submission of proofs). If the proof is not submitted continuously and in a timely manner within the submission cycle, some tokens will be deducted by the system.

The process of generating proof is very similar to that of replicating proof, except that the input of space-time proof is the previously generated proof as input parameter, which ensures the continuity of proof generation. See the white paper for details.

4. Proof of Spacetime and Consensus Mechanism

The consensus mechanism of blockchain is one of the core elements of blockchain projects. Filecoin 's consensus mechanism reaches consensus based on proof of time and space. This consensus mechanism is as simple and clear as POW . The following is an example from Star Continent Mingge:

(1) Suppose there are two miners in the network: A and B. Miner A seals two sectors full of data at a certain moment and keeps submitting time-space proofs. Miner B seals four sectors full of data at this moment and keeps submitting time-space proofs .

(2) At this time, the effective storage of the entire network is 6 sectors

(3) The probability of miner A being elected as the leader node is 2/6

(4) The probability of miner B being elected as the leader node is 4/6

(5) That is, the probability of a miner being elected as a leader node at a certain moment = sealed sector capacity / total sealed sector capacity of the entire network

5. Miners, Miner Actors and Mining Machines

According to the demo released by the Filecoin team , the details of Filecoin mining are quite different from those of digital currencies such as BTC and ETH . There are miners and miner actors. So what is the relationship between miners, miner actors and mining machines? Interstellar Continent Mingge’s understanding is as follows for reference by all miners:

1. Miners

Miner: It can be understood as an address and key pair similar to BTC and ETH , which is needed when issuing orders and completing orders. The mining income also belongs to the miner's address.

2. Miner Actor

Miner actor: Each miner can have multiple miner actors. When issuing orders, the miner actor is the unit (see the demo video translated by Star Continent for details, watch Juan's demonstration of the miner's bid carefully, he copied the address of the miner actor). Each miner actor also has an address, and the storage of multiple miner actors of the miner can be accumulated (Ming Ge is just guessing).

3. Mining Machine

Each miner actor corresponds to a physical mining machine.

6. About Filecoin Mining Pool

According to our understanding, Filecoin may not be able to realize the traditional virtual mining pool like BTC and ETH for the following reasons:

(1) Miners need to use a key pair when issuing orders and generating proofs. It is very dangerous for the mining pool service provider to tell the miners the secret key (unlike traditional mining pools where only a mining pool address needs to be configured). Of course, this may be achieved through encryption, but the specific possibility cannot be determined at present.

(2) Miners need to continuously submit storage proofs. If some miners in the mining pool are unable to submit proofs in a timely manner, the deducted collateral will be shared by all miners, which is very unfair to high-quality miners.

(3) Mining pool service providers need to provide a very large token mortgage, because the storage of each miner is constantly changing, and it is unlikely that miners can directly pay all mortgage fees, which is also unfair.

In summary, the feasibility of the Filecoin virtual mining pool is questionable, but it does not mean that there will be no innovative methods, such as setting up a physical mining pool, and having the mining farm or mining pool service provider uniformly manage the mining machines. For miners, the same effect of joining the mining pool can be achieved.