Storage

Tailor storage to meet high-quality data requirements

In EpiK Protocol, AI data is formatted in the graph database log, including CREATE, UPDATE, and DELETE statements. The format allows us to define the schema of diverse data in a unified design and makes data updatable, traceable, and tamperproof. Anyone can recover the graph database by replaying the ordered statements in the log files.

AI data is fundamental to AI implementation. If anyone could upload data at will, the quality of the data could not be guaranteed. To avoid this situation, therefore, we have incorporated in our protocol the role of elected domain experts. Only data Domain Experts have approved and accepted can be uploaded to the EpiK Protocol network. Data uploaded by different domain experts are independent. As such, data uploaded by one Domain Expert does not influence nor impact data uploaded but another Domain Expert.

High-quality requirements limit the rate at which data can grow and prevent an arms race, such as the use of meaningless data to increase mining power in the Filecoin network endlessly. The limited amount of new data added daily will effectively narrow the mining power gap between home computers and high-performance professional computers.

Lowering mining node deployment costs brings a series of benefits to the network. When node costs decrease, the number of nodes increases. When the number of nodes increases, the number of global backups of the same data can also increase. As the number of backups grows, the penalty for a single node downtime can be reduced or even eliminated. When the penalty becomes smaller, it will further reduce node costs.

In addition to this, we have added a series of designs around storage to reduce node costs even further.

• The sector size is fixed to 8M so that nodes without GPUs can also participate in mining.

• The first 100 nodes that complete the file storage will get double mining power for a new file, thereby incentivizing file backups across the network.

• The penalty for downtime nodes is canceled to reduce operation and maintenance costs and introduce home computer mining.

The facts do confirm our hypothesis. At present, the number of mining nodes in the EpiK Protocol network exceeds 50,000. Most of them are not deployed in professional computer rooms but are placed in users' homes. The average number of backups for each file on the entire network exceeds 10,000, and the impact of a single node downtime on file availability is almost zero.