If you choose to set the read mode to staging, the agent creates a staging file in the directory that you specify in the mapping properties. You need to specify a staging directory on the Secure Agent machine that has sufficient disk space corresponding to the volume of data that you want to process.

When you run the mapping, the UNLOAD command loads the data from the Amazon Redshift table to the Amazon S3 bucket, and then the agent loads the data from S3 to the staging directory on the agent machine. After the agent completes the read operation, the staging files are deleted.

When you set the read mode to direct, Data Integration directly accesses and retrieves the data from Amazon Redshift without staging the data.

When you configure the direct read mode, you can also specify the fetch size to determine the number of rows of data to retrieve from the Redshift source at a given time. The advanced source attributes applicable for S3 staging in the mapping properties and the staging optimization property in the agent properties are not required for the direct mode.

For the list of advanced source attributes that apply for direct and staging read modes, see Amazon Redshift V2 sources in mappings.