Distributed storage system and data processing method

A data processing method implemented a distributed storage system having a plurality of storage devices, comprising:

• sending, by a first artificial intelligence (AI) apparatus of a first storage device in the plurality of storage devices, a data obtaining request to a first processor of the first storage device to obtain service data, wherein the first storage device of the distributed storage system comprises the first processor, a first hard disk, and the first AI apparatus, wherein the first AI apparatus communicates with the first processor through a high-speed interconnect network;
• sending, by the first processor, metadata of the service data to the first AI apparatus in response to the data obtaining request, wherein the metadata indicates an address of the service data:
• when the metadata indicates that the service data is located in a second storage device in the plurality of storage devices, sending, by the first AI apparatus, a second data access request to the second storage device, wherein the second data access request is a remote direct memory access (RDMA) request, and wherein the second data access request comprises the metadata;
• receiving, by the first processor, the service data sent by the second storage device in response to the second data access request;
• storing, by the first processor, the service data in the first hard disk;
• and
• performing, by the first processor, AI computing on the service data.