Query Plan Distribution and Optimization
When dealing with millions of rows from different sources, or when short response time is important, being able to minimize the amount of data to be processed can decrease the actual response time a great deal. The following are some of the techniques implemented by Denodo to minimize response time and optimize query plan execution:
- Process & Function Delegation to the Sources
When a query is executed against Denodo, or a virtualized data service is invoked, the request is preprocessed and transformed into a query plan execution using relational algebra techniques. Denodo employs the efficient technique of delegating operations to the sources so that much of the processing is done by the sources.
- Cost-based Join Strategy Selection
Another key aspect of query plan cost is Denodo's ability to choose the most appropriate strategy for join operations. Technically speaking, this is an expensive operation in terms of performance and much has been written about it in the research world. Denodo automatically tries to use the most appropriate strategy in each case based on internal cost information, but also allows advanced administrators to control the specific strategies being used.
- Dynamic Choice of Strategies
Denodo usually chooses the most efficient operator strategy in creating a view. However, Denodo allows the administrator to override this setting and to customize the operator strategy used for a particular view. This allows the Data Stewards and Engineers to make sure they are always providing the most efficient query plan.
