On In-Memory Cubes

Share on LinkedInTweet about this on TwitterShare on FacebookShare on Google+

OLAP (On-Line Analytical Process) data storage solutions have been available from the 1990s to solve the problem of keeping tabs on KPIs (Key Performance Indicators). OLAP cubes are more efficient than relational tables because only dimensions and aggregated measures for each dimension are stored in the cube. Running queries against the transaction oriented OLTP (On-Line Transaction Process) tables often require too much time and too many resources. Typically the OLAP cubes are updated daily or weekly as the schedule permits and users can interact with this historical data to understand past performance and be made aware of trends to predict future performance and be able to react quickly to KPI exceptions both good and bad.  For example, in the fashion industry buyers have a very short window of 1 or 2 days to decide whether to reorder if sales are good or cancel orders if sales of a product are lower than expected.

There is a major installed base of OLAP products but there is less interest in OLAP tools now as customers realize that these tools are not designed for transactional systems.  Since these systems experience frequent updates and large data volumes, the OLAP cubes extracted from them are often out-of-date.

A better solution has been developed and was enabled by 64-bit processors and vast amounts of very low cost memory.  People realized that by directly accessing transactional relational databases and creating in-memory cubes they could still get great performance, interact with current data, answer questions about their data and monitor KPIs without an expensive OLAP product.

Leveraging this concept, JReport allows users to view their KPIs and do real-time analysis of their data without extracting the data into traditional OLAP cubes.  Instead, a cube structure is simulated in memory directly using SQL queries on relational databases.

With JReport you can use the same dashboards and reports to directly access the relational database, access cached datasets from disk, access in-memory cubes containing just aggregated data, or optionally, access all of the detail data for drilling in to root cause problems. JReport is a full BI platform that provides a common tool for building ad hoc reports, dashboards, and interactive analysis using new components or reusing existing components.

To hear an in-depth discussion on how in-memory cubes increases performance for data analysis, attend our session on “Performance Tuning for Fast Reporting and Analysis” at JReport Summit 2013, our online conference on April 2.

Share on LinkedInTweet about this on TwitterShare on FacebookShare on Google+

Leave a Reply

Your email address will not be published. Required fields are marked *

Test * Time limit is exhausted. Please reload CAPTCHA.