Doris Cluster Architecture (Nodes, Shards, Data Replication, etc.)

Node (Node)

In a Doris cluster, a node represents a machine, which can be a single machine or a combination of multiple machines. Each node has its own data storage and can handle read/write requests from other nodes.

Shard (Shard)

A shard is the smallest unit in a Doris cluster. One shard corresponds to one table, which can be divided into multiple replicas and distributed across different nodes. Shards on each node are independent and can also be divided into more replicas.

Data Replication (Replication)

Data replication is a key feature in a Doris cluster that ensures all nodes have the latest complete data replica. When a new node is added to the cluster, data replication must be performed to ensure the new node also has complete data.

Combination of Nodes and Shards

In a Doris cluster, a node can simultaneously process multiple shards, while each shard corresponds to a node. This can achieve high availability and high performance.

Data Read/Write Requests

When a client sends a data read/write request to a Doris cluster, it will first attempt to obtain the required data from the local cache. If not found, it needs to send read/write requests to other nodes, which may require waiting for a period of time before requesting again.

Cluster Scaling

A Doris cluster can be scaled up or down according to needs. Adding new nodes will improve cluster performance, while removing old nodes will reduce cluster performance.