Get A POC
Gartner's key advice to would-be DW buyers is to perform proof-of-concept (POC) assessments among a few finalist candidates. These tests should be done at the customers site with the customer's data and with as many data sources, users and simulated workloads as possible.
Most buyers heed this advice, says Gartner, but some vendors make things difficult. IBM, for example, has been pickier about participating in POCs while Oracle avoids on-site POCs entirely, pushing customers to perform such tests at one of nine international Exadata test sites. Given Oracle's market share, you could imagine on-site POC demands getting overwhelming.
What customers increasingly need is the ability to handle mixed workloads and to optimize performance for specific needs. Gartner encourages thorough assessment across six capabilities: bulk/batch loading, reporting, online analytical processing (OLAP), real-time/continuous load, data mining and operational BI.
(click image for larger view)
To optimize these workloads, hardware management for input/output, disk store and CPU/memory balancing are now included "as a matter of course" in DW platforms, Gartner notes, but it's important to explore capabilities and flexibility.
Storage optimization and compression are also receiving a lot more attention. EMC has stepped forward on storage management and Oracle and the column-store vendors on compression.
Vendors are putting a lot of performance and technology claims out there these days, but these differentiators "are not necessarily significant to the use case," Gartner cautions. That's why POC with your data, your workloads and your user expectations are vital to success.
A complete copy of Gartner's Magic Quadrant report is available from this link at the Sybase Web site.
