Data distribution method, device, computing node and storage system in storage system

Abstract

The invention discloses a data distribution method, device in a storage system, calculation nodes and the storage system belonging to the computer technical field. The storage system comprises multiple calculation nodes and multiple storage nodes. The method comprises that: a target calculation node receives an input output I / O request sent by an external node, the target calculation node is determined by the external node from the multiple calculation nodes according to a first algorithm; the target calculation node determines a target storage node from the multiple storage nodes according to a second algorithm; and the target calculation node sends an I / O request to the target storage node. According to the method provided in the invention, the calculation function and the storage function of the nodes are deployed in separation; therefore, the storage system comprises the multiple calculation nodes and the multiple storage nodes; the calculation performance and the storage performance are not restricted and influenced mutually; the calculation performance and the storage performance of the system are configured flexibly according to demands; the calculation performance is independent of the storage performance; and the software and hardware are decoupled.

Description

technical field [0001] The present invention relates to the field of computer technology, in particular to a data distribution method, device, computing node and storage system. Background technique [0002] In a distributed storage system, nodes are generally formed into a DHT (Distributed Hash Table, Distributed Hash Table) ring, each node is responsible for managing a section of the ring (partition), each I / O (Input / Output , input and output) requests are stored on the node corresponding to the area where the hash value is located. In order to ensure data consistency, primary-secondary algorithms are commonly used in distributed storage systems to store data. The primary-secondary algorithm is a master-slave algorithm, which is generally used to store multiple copies of data. In this algorithm, the external node first sends the I / O request to the master node, and then the master node synchronizes the I / O request to the remaining nodes to ensure the consistency between t...

AI Technical Summary

Problems solved by technology

[0005] Because the storage function and the control function of the node are integrated, the control function is strongly dependent on the storage function, making it impossible to decouple the software and hardware. The control performance and storage performance are mutually restricted and influenced by each other. It affects the control function. When the control function fails, it will also seriously affect the storage function, causing the performance of the entire distributed system to be damaged.

Images