Highlights of this year's data warehouse DBMS Magic Quadrant include:
- Teradata is #1, Oracle is #2, and IBM is #3, with the first two if anything slightly extending their leads. (in 2006, IBM was #2.)
- Netezza has been given a nice upwards (actually, more rightwards) bump and is now a clear #4.
- Microsoft is treading water at a clear #5.
- Greenplum and Sybase have slid back some, but depending on which dimension you weight more heavily are somewhere in the #6-8 range.
- HP joins newly, as the other #6-8 competitor, a little behind Sybase.
- Vertica joins as a first-timer, as a clear #9.
- Kognitio and SAND are next, with hefty gains in "ability to execute", both leapfrogging Sun/MySQL.
- Ingres, iLLuminate, and 1010data straggle in at the bottom, all of them new (at least versus 2006-2007).
I don't really have a lot of quarrel with the "completeness of vision" rankings. As I see it, important attributes of a data warehouse DBMS "vision" would include:
- A performance story across at least a reasonable range of workloads.
- Either a clear hardware architecture story, or else a clear story as to why hardware architecture is relatively unimportant.
- SQL 2003 and further features in integrated analytics.
- Reasonable OLTP-like features, from the basics - ACID compliance! - to manageability, high availability and fast-enough update/load.
- Good compatibility with third-party products.
Gartner's rankings are not ridiculous by those standards. Aster would surely have ranked high, but obviously they did not meet the confirmed-sale requirements for inclusion.
So what about Gartner's "ability to execute" rankings? These are approximately:
- Teradata at #1
- Oracle and IBM tied at #2-3
- HP, Sybase, Microsoft, and Netezza tied at #4-7
- Greenplum at #8, Vertica at #9, and everybody else trailing after
That looks like it's basically a measure of revenue, blending overall corporate and data-warehouse-DBMS-specific figures in some way, adjusted for who can deploy the most credible-sounding executive who appears to simultaneously have his - I use the male pronoun deliberately - finger on development and revenue-generation alike.
Frankly, I think it's that dimension that makes Gartner Magic Quadrants well-nigh meaningless. If you asked me in which vendor's execution-on-vision I had the most confidence, I'd stammer around unless I felt free to reframe the question and shoot back "Which PART of the vision?" If you want to deploy a 1 terabyte data warehouse with a highly diverse workload - well, Oracle, IBM, Teradata, and to a lesser extent Microsoft have been doing that for years, and they deserve to be atop the ability-to-execute charts, with Netezza perhaps not far behind. If you want to run fast queries on cheap hardware on 200 GB of data, Sybase IQ is a proven market leader. If you want a cheap 100 TB data warehouse that will soon scale to over a petabyte, Oracle's great achievements in other areas of DBMS and its clever Exadata ideas suffice merely to put it on a par with those smaller vendors that have actually deployed a few such systems each.
When selecting a database management system for analytic processing, confine yourself to those vendors whose products can, today, do everything you're likely to need for the next few years. Further require that they be on track to soon deliver most of what you seriously want over that time period. And throw the Gartner MQ into the nearest bit bucket, before it confuses your evaluation cycle irredeemably.Gartner's latest Magic Quadrant for data warehouse DBMSs was published late last year. Thankfully, vendors don't seem to be taking it as seriously as usual, so I didn't immediately hear about it... Highlights of this year's Quadrant include...