Client Performance and Optimization Settings
The client performance and optimization settings are available from the Options tab of the Stream Properties dialog box. To display these options, choose the following from the client menu.
You can use the Optimization settings to optimize stream performance. Note that the performance and optimization settings on IBM® SPSS® Modeler Server (if used) override any equivalent settings in the client. If these settings are disabled in the server, then the client cannot enable them. But if they are enabled in the server, the client can choose to disable them.
If connectivity is enabled, you see the option Server Enablement in the License Status tab.
See Connecting to IBM SPSS Modeler Server for more information.
Enable stream rewriting. Select this option to enable stream rewriting in IBM SPSS Modeler. Four types of rewriting are available, and you can select one or more of them. Stream rewriting reorders the nodes in a stream behind the scenes for more efficient operation, without altering stream semantics.
- Optimize SQL generation. This option enables nodes to be reordered within the stream so that more operations can be pushed back using SQL generation for execution in the database. When it finds a node that cannot be rendered into SQL, the optimizer will look ahead to see if there are any downstream nodes that can be rendered into SQL and safely moved in front of the problem node without affecting the stream semantics. Not only can the database perform operations more efficiently than IBM SPSS Modeler, but such pushbacks act to reduce the size of the data set that is returned to IBM SPSS Modeler for processing. This, in turn, can reduce network traffic and speed stream operations. Note that the Generate SQL check box must be selected for SQL optimization to have any effect.
- Optimize CLEM expression. This option enables the optimizer to search for CLEM expressions that can be preprocessed before the stream is run, in order to increase the processing speed. As a simple example, if you have an expression such as log(salary), the optimizer would calculate the actual salary value and pass that on for processing. This can be used both to improve SQL pushback and IBM SPSS Modeler Server performance.
- Optimize syntax execution. This method of stream rewriting increases the efficiency of operations that incorporate more than one node containing IBM SPSS Statistics syntax. Optimization is achieved by combining the syntax commands into a single operation, instead of running each as a separate operation.
- Optimize other execution. This method of stream rewriting increases the efficiency of operations that cannot be delegated to the database. Optimization is achieved by reducing the amount of data in the stream as early as possible. While maintaining data integrity, the stream is rewritten to push operations closer to the data source, thus reducing data downstream for costly operations, such as joins.
Enable parallel processing. When running on a computer with multiple processors, this option allows the system to balance the load across those processors, which may result in faster performance. Use of multiple nodes or use of the following individual nodes may benefit from parallel processing: C5.0, Merge (by key), Sort, Bin (rank and tile methods), and Aggregate (using one or more key fields).
Generate SQL. Select this option to enable SQL generation, allowing stream operations to be pushed back to the database by using SQL code to generate execution processes, which may improve performance. To further improve performance, Optimize SQL generation can also be selected to maximize the number of operations pushed back to the database. When operations for a node have been pushed back to the database, the node will be highlighted in purple when the stream is run.
- Database caching. For streams that generate SQL to be executed in the database, data can be cached midstream to a temporary table in the database rather than to the file system. When combined with SQL optimization, this may result in significant gains in performance. For example, the output from a stream that merges multiple tables to create a data mining view may be cached and reused as needed. With database caching enabled, simply right-click any nonterminal node to cache data at that point, and the cache is automatically created directly in the database the next time the stream is run. This allows SQL to be generated for downstream nodes, further improving performance. Alternatively, this option can be disabled if needed, such as when policies or permissions preclude data being written to the database. If database caching or SQL optimization is not enabled, the cache will be written to the file system instead. See the topic Caching options for nodes for more information.
- Use relaxed conversion. This option enables the conversion of data from either strings to numbers, or numbers to strings, if stored in a suitable format. For example, if the data is kept in the database as a string, but actually contains a meaningful number, the data can be converted for use when the pushback occurs.