Architecture Description

IBM® SPSS® Modeler Server uses a three-tier, distributed architecture. Software operations are shared between the client and the server computers. The advantages of installing and using IBM SPSS Modeler Server (versus the standalone IBM SPSS Modeler), especially when dealing with large data sets, are numerous:

IBM SPSS Modeler Server can run on UNIX, in addition to Windows, allowing more flexibility in deciding where to install it. On any platform, you can dedicate a faster, larger server computer to data mining processes.
IBM SPSS Modeler Server is optimized for fast performance. When operations cannot be pushed into the database, IBM SPSS Modeler Server stores the intermediate results as temporary files on disk rather than in RAM. Because servers usually have significant disk space available, IBM SPSS Modeler Server can perform sort, merge, and aggregation operations on very large data sets.
Using the client-server architecture, you can centralize data-mining processes in your organization. Centralization can help to formalize the role of data mining in your business processes.
Using administrator tools like the IBM SPSS Modeler Administration Console (included with IBM SPSS Deployment Manager) and IBM SPSS Collaboration and Deployment Services (sold separately), you can monitor data mining processes, ensuring that adequate computing resources are available. With IBM SPSS Collaboration and Deployment Services you can automate certain data mining tasks, manage access to data models, and share results across your organization.

The components of IBM SPSS Modeler's distributed architecture are shown in the "IBM SPSS Modeler Server Architecture" graphic.

IBM SPSS Modeler. The client software is installed on the end user's computer. It provides the user interface and displays the data mining results. The client is a complete installation of IBM SPSS Modeler software, but when it is connected to IBM SPSS Modeler Server for distributed analysis, its execution engine is inactive. The IBM SPSS Modeler runs on Windows operating systems only.
IBM SPSS Modeler Server. The server software installed on a server computer, with network connectivity to both the IBM SPSS Modeler(s) and the database. IBM SPSS Modeler Server runs as a service (on Windows) or a daemon process (on UNIX), waiting for clients to connect. It handles the execution of streams and scripts created using the IBM SPSS Modeler.
Database server. The database server could be a live data warehouse (for example, Oracle on a large UNIX server) or, to reduce impact on other operational systems, a data mart on a local/departmental server (for example, SQL Server on Windows).

IBM SPSS Modeler Server Architecture

With the distributed architecture, most of the processing occurs on the server computer. When the end user executes a stream, IBM SPSS Modeler sends a description of the stream to the server. The server determines which operations can be executed in SQL and creates the appropriate queries. These queries are executed in the database, and the resulting data are passed to the server for any processing that cannot be expressed using SQL. Once the processing is complete, only the relevant results are passed back to the client.

If necessary, IBM SPSS Modeler Server can execute all IBM SPSS Modeler operations outside of the database. It automatically balances its use of RAM and disk memory to hold data for manipulation. This process makes IBM SPSS Modeler Server fully compatible with flat files.

Load balancing is also available by using a cluster of servers for processing. Clustering is available starting in IBM SPSS Collaboration and Deployment Services 3.5 through the Coordinator of Processes plug-in. See the topic Load Balancing with Server Clusters for more information. You can connect to a server or cluster managed in the Coordinator of Processes directly through IBM SPSS Modeler's Server Login dialog. See the topic Connecting to IBM SPSS Modeler Server for more information.

Standalone Client

IBM SPSS Modeler may also be configured to run as a self-contained desktop application, shown in the graphic below. See IBM SPSS Modeler Support for more information.