Types of queries that are not accelerated

There are characteristics of queries that either will not benefit from being sent to the accelerator server, or will not be considered for acceleration.

Queries that will not benefit from being accelerated

Queries that only refer to a single, small dimension table do not benefit from being sent to the accelerator server for processing as much as queries that also refer to a fact table.

Queries that return a large result set should be processed by the database server to avoid the overhead of sending the large result set from the accelerator server over the connection to the database server. If a query returns millions of rows, the total response time of the query is influenced by the maximum transfer rate of the connection. For example, the following query might return a very large result set:

SELECT * FROM fact_table ORDER BY sales_date;

Queries that search only a small number of rows of data should be processed by the database server to avoid the overhead of sending the query to the accelerator server.

Queries that are not considered for acceleration

There are some queries that will not be processed by the accelerator server.

Queries that would change the data cannot be processed by the accelerator server and must be processed by the database server. The data in the accelerator server is a snapshot view of the data and is read only. There is no mechanism to change the data in the data marts and replicate those changes back to the source database server.

Other queries that are not processed by the accelerator server include queries that contain INSERT, UPDATE, or DELETE statements, queries that contain subqueries, and other OLTP queries.