Ovum Decision Matrix: Selecting an Analytic Database, 2013-14

Melbourne, 13 February, 2014 – Ovum identifies the market leaders for analytic database solutions are IBM PureData, SAP HANA, Teradata Aster Discovery Platform, and HP Vertica. However, Kognitio, SAP Sybase IQ, and ParAccel are strong challengers that have the potential to break into the leader category within the next year.

The emergence of purpose built and workload-optimised analytic databases over the past two to three years has shaken up the previously commoditized database market. Relational databases have started to make headlines due to the advent of Big Data and alternative “NoSQL” database infrastructures such as Hadoop.

In Ovum’s latest Decision Matrix: Selecting an Analytic Database, 2013-14* the independent global analyst firm reveals that an increasing number of organisations are now revisiting their current analytics strategies and are looking to implement specialised analytic databases that promise to keep pace by delivering more advanced analytics in bigger and faster data environments.

“The business case of purchasing and implementing an analytic database is becoming clearer. Traditional data warehousing architectures are struggling to manage larger data volumes, handle new type of human and machine generated data, deliver rapid query response and devise more sophisticated analytics,” say Surya Mukherjee, senior analyst at Ovum.

To address these new requirements, vendors have incorporated certain core design principles in both their software and hardware for improving analytic processing performance. Some of the strategies include conventional query optimization, smart index management, pre-calculated views, in-memory processing and columnar storage for software. While, hardware strategies that help speed and improve query performance include using architectures such as MPP, in-memory engines, interconnects and pre-configured appliances.

Ovum’s research finds that the use of parallelization is the most commonly applied strategy where all solutions we evaluated support parallel loading and parallel querying of data. Support for in-memory processing technology is also a common thread across the group. However, the level to which solutions use in-memory varies widely across vendor solutions.

Ovum believes that all these strategies are useful in specific analytic scenarios; some are effective for organisations that need consistent real-time responses on very large data sets above others.

“An ideal – for performance purposes – would of course be to put all data into memory. However, despite the falling cost-versus-performance, it is still more expensive than partial disk and memory platforms, and the cost-versus-performance differential influences enterprises’ selection processes,” Mukherjee comments.

The ODM reveals that all databases covered in this report provide strong support for advanced analytics, either in-database, through UDFs (user-defined functions), or by supporting parallelized versions of R open source statistical software.

Enterprises are advised to explore the depths of the predictive libraries of the vendors they shortlist – these can range from scores to thousands of pre-built functions. Enterprises must also be aware that exporting existing statistical models directly to a newly deployed analytic database might entail modification of existing code and additional expenditure on services.

“A point to note in our research is that all the vendors who participated, support fast deployment of cloud and appliances. The approaches to cloud and appliances do vary, with some such solutions available only on pre-configured appliances or private clouds, while others are available as commodity hardware. Cloud deployment will undoubtedly be a key area of development for all database vendors in the medium term,” concludes Mukherjee.