In-memory analytics is an emerging business intelligence trend. This enables the analytics to be performed in RAM, therefore avoiding the need for time-consuming hard-disk i/o tasks. This has a trememndous potential of quantum jump in BI performance. The announcement of Gemini project from microsoft on October 6,2008 has triggered the competetion in the area of BI.

Broadly speaking, using in-memory analytics the entire reference data for running your queries, exists in the RAM and not on the hard-disk. The main bottle-neck for the Business Intelligence performance has been the time taken to access the hard-disk and I/O activities. With in-memory analytics, this barrier is overcome to a great extent.

In-memory analytics trend is catching up in line with the advancement and cost-reduction in hard-ware technology, for example

•     Whereas a single 32 bit CPU can manage only 2-3 GB data in RAM, a 64 bit CPU can manage over 100GB.
•     The cost of RAM has moved down significantly.   

Benifits of in-memory analytics:

Faster response time.

RAM access speeds can be million times faster than that of a hard-disk. Unlike a hard-disk where the access is done through memory blocks, in RAM the access can be done to a specific point.

Consistent response time

Due to faster access speeds, the response time between "light"and "heavy" query is not significant. In RAM it does not make much difference if a query has to access 10 or 100 tables.

Less data-size

Less data-size is required for storing the same data. The reason is that for in-memory analytics you don't need to have too much pre-calculated or pre-aggregated data. The pre-calculation and pre-aggregation is done to reduce the data access time taken for a disk-based system. With RAM instant access capability, one need not create huge indexes and pre-calculations. In RAM one can afford to have entire table scans almost instantaneously.

Reduced development and performance tuning effort :

For hard-disk based analytics, Business Intelligence designs spend huge time in terms of deciding the balance between the data size v/s the response time improvements. A lot of effort goes into allocating the caching space and also how to manage caching across the application servers and database layer. While RAM level caching (maintaining the data in RAM till it is overwritten by the new data) is almost automatic for major DBMS solution, but allocation of RAM cache across application severs and DBMS layer is still managed manually.

Examples of in-memoery analytics in market:

• SAP Business Intelligence Accelerator and Applix TM1.
• TM1 Applix.
• Qlikview from Qlik Tech.
• Panoratio
• Microsoft Gemini(to be released in 2010)

Current(2008) market penetration of in-memory analytics has been from 1% to 5% of the total market. This share is bound to rise fast in the coming days.