PowerPlay is the multidimensional analysis component of the Cognos BI Series 7 suite. The suite contains an integrated set of products that provide end-to-end business intelligence solutions. It includes an extract, transform, load (ETL) tool; executive-level dashboards and scorecards; multidimensional analysis (for online analytic processing, or OLAP); managed query and enterprise reporting; and alerting and event detection.

PowerPlay, known for its ease of use, provides two ways to view multidimensional data: Explorer, for ad hoc analysis, and Reporter, with more formatting and calculation options. Version 2, released in March, provides functional parity between the Explorer PC windows and Web user interfaces. Web Explorer uses HTML and dynamic HTML, so users don't have to download and install Java applets or propriety plug-ins to their PCs — providing a zero footprint that facilitates easier intranet or extranet deployment. Other enhancements include projecting future data based on history and support for PowerCubes (Cognos's multidimensional cubes) with increased capacity.

PowerPlay Architecture

The PowerPlay architecture is multitier. (See Figure 1.) It has one common application server (the PowerPlay Enterprise Server, or PPES) and three clients (Web, Windows, and Excel). Server components can reside on one or separate machines, and the results of client queries that PPES processes are stored in a shared server cache. When possible, subsequent queries are served from this cache, which speeds up query time.

Client connections to PPES are stateless; after PPES processes a request, it ends the session. Doing so prevents overall performance degradation from users who are idling or simply reading information. This feature enables greater scalability by more efficiently using a given bandwidth.

PowerPlay can access information in its own PowerCubes, as well as information in third-party OLAP servers, such as Microsoft Analysis Services, SAP BW, Hyperion Essbase, and IBM OLAP for DB2. This OLAP server agnosticism allows companies to leverage existing investments in technology infrastructure and skill sets.

Transform Information into Cubes

You use PowerPlay Transformer to design and build PowerCubes. Information can be loaded from multiple sources including Oracle, DB2, and files with formats such as fixed or delimited text files and Excel.

PowerCubes consist of measures and dimensions that have hierarchies and levels. Individual dimension members are called categories. PowerCubes are stored in a propriety, compressed, bitmapped index structure where individual data cells are further compressed. Data aggregation from the lowest-level input categories to higher levels in the hierarchies is intelligently automated so that only some of the higher levels are calculated and stored, providing fast response time. Levels between the high and low are dynamically aggregated at run-time.

This degree of automatic compression and aggregation obviates design decisions necessary in OLAP servers such as Oracle Express, Oracle9i OLAP, and Hyperion Essbase; instead, designers can focus on building cubes that meet business requirements. These decisions include identifying dense and sparse dimensions, physical ordering of dimensions, and which hierarchy levels to aggregate — all of which can affect performance considerably.

Series 7 version 2 supports PowerCubes with up to 2 million categories (calculated by adding the total number of categories along each dimension) and up to 1 billion rows of consolidated data. PowerCubes can be partitioned (stored in different physical files), which increases data capacity and, like compression, is automatic. Time-based partitioning, new in version 2, provides faster cube updates where data is loaded for a specific time period by incrementally updating separate time-based PowerCubes (such as monthly) instead of processing all time periods. To the user, both types of partitions appear as a single PowerCube.

Measures must be numeric (eight-byte reals and two- or four-byte integer types) and can be stored or calculated, have different dimensions, and contain information at different levels. For example, budget would typically be loaded at a higher level than actuals or with fewer dimensions. Transformer can allocate data down to lower levels (for example, from year down to month). Variance (actual minus budget) would be a calculated measure. Measures hierarchies are supported. Higher-level measures can be simple roll-ups or calculations with more sophisticated logic. In this way, chart-of-account dimensions can be implemented to support financial analysis. However, PowerPlay is read-only and wouldn't be a suitable tool for collaborative read/write budgeting and planning applications.

PowerPlay supports multiple hierarchies (alternate drill-down paths) along the same dimension, providing that they have at least one hierarchy level in common and the categories for this level are unique. For example, some companies would want to view the organization dimension by both geography and legal entity. However, having geography and legal entity as different dimensions would provide more reporting flexibility. PowerPlay doesn't support the concept of attributes — the ability to group categories (for example, grouping customers by ZIP Code or grouping product by size and color). Alternatively, each attribute can be a separate dimension or accessed by drill-through. (More on this later.) There's no hard limit to the number of dimensions in a PowerCube, but in general, the greater the number of dimensions, the more difficult it is for users to navigate. For applications that require a large number of dimensions, PowerPlay allows drill-across to different cubes as long as they have at least one dimension in common. For example, you could identify the least-profitable customers in the finance cube and drill to the sales cube for more information about these customers, such as product mix or salesperson.