Insert Performance

A INSERT INTO statement is processed as follows:

• Parse the statement to create an abstract syntax tree

• Do some basic semantic validation

• Plan the operation

• Execute the operation

CrateDB calculates the shard ID for every row to be inserted when executing the operation. Insert requests are then grouped and sent to the nodes that hold each primary shard.

You can reduce the processing overhead by either eliminating the needless repetition of some steps or by reducing the work needed to be done by one or more steps.

This section of the guide will show you how.

Table of contents

• Insert Methods
  • Statement types
  • Client approaches
  • Testing
• Bulk Inserts
  • Configure your tables
  • Import the data
  • Split your tables into partitions
• Parallel Inserts
  • Example
  • Testing
• Configuration Tuning for Inserts
  • Hardware
  • Software
• Testing Insert Performance
  • Test data
  • Running a test