Term

Definition

Management Node

The management node maintains the Cluster Configuration and hands out the configuration to the data nodes, SQL nodes, and Application Nodes. This functionality is incorporated in [CZ].

Data Nodes

The Data Storage which is incorporated in the Data Repository.

MySQL Server

The interface to the data nodes for creating/modifying schema and querying data. This functionality is incorporated in [CZ].

ClusterControl Server

This server is included in the Management Node and provides the other computers in the cluster with their respective configurations. This functionality is incorporated in [CZ].

Local checkpoint

A local checkpoint is when a copy of the database is written to disk. Each data node backup its partitions. If you have two data nodes, each of the nodes will backup its own half of the database. However, if one of the data nodes crash, the other data node will backup the entire database.

There will always be at least one full local checkpoint of the database written to disk. When one has been completely written, the next will start overwriting the oldest checkpoint

redo log

A redo log containing all the changes to the database will always be written as well. The redo log contains all the operations that have been made from at least two check points back. The changes stored from the oldest checkpoint will not be deleted until all the changes for a third checkpoint have been saved.

Partition

The database is divided into partitions where each data node is in charge of one partition each. The data is stored in a partition based on a hash made of the data so that each partition will always contain an equal share of the data in the database. If you have two replicas, then each node will have one partition which it is in charge of, and one secondary partition which is fully replicated. If the data node which is in charge of the partition crashes, the second data node with the replication will take charge of that partition as well.

If you have two data nodes, then data node 1 will be in charge of partition 0 and have partition 1 as a replica, while data node 2 will be in charge of partition 1 and have partition 0 as a replica.