February, 2011 edit: I’ve now commented on Gartner’s 2010 Data Warehouse Database Management System Magic Quadrant as well.
At intervals of a little over a year, Gartner Group publishes a Data Warehouse Database Management System Magic Quadrant. Gartner’s 2009 data warehouse DBMS Magic Quadrant — actually, January 2010 — is now out.* For many reasons, including those I noted in my comments on Gartner’s 2008 Data Warehouse DBMS Magic Quadrant, the Gartner quadrant pictures are a bad use of good research. Rather than rehash that this year, I’ll merely call out some points in the surrounding commentary that I find interesting or just plain strange.
*Links to Gartner Magic Quadrants commonly break, but that one worked at the time of this posting.
- Gartner thinks that data warehouse appliances are on the rise, due to their simplicity.
- Gartner correctly says that Teradata has been a data warehouse appliance vendor from the getgo.
- Gartner characterizes IBM as being an appliance vendor as well.
- Gartner suggests that HP is having trouble living up to its technical promises for Neoview.
- Gartner further suggests — no surprise here — that HP Neoview has had very few new customers past its initial wave.
- Gartner notes IBM’s difficulties in selling data warehouse installations of DB2, despite what on paper is great-sounding technology.
- Gartner says — also no surprise — that illuminate “has seen little success in North America since opening its first office in the U.S. over two years ago.”
- Ingres has evidently gotten a few BI-centric “appliance” deals, e.g. with Jaspersoft. But basically Ingres isn’t doing well in data warehousing.
- Gartner does say Ingres has “the strongest open-source DBMS offering for data warehousing.” Being very literal about “open source,” that’s a defensible claim — but it’s pretty irrelevant in a world where Greenplum Single-Node Edition can be had for free. It also waves away all the data mart use cases in which Infobright Community Edition shines.
- Gartner says that Netezza is working out as a “complex workload” enterprise data warehouse provider, according to reference checks, in addition to its established success in data mart scenarios.
- Gartner says Oracle’s offering has finally become “accepted” in the market for databases >50 TB. I guess I can live with that fairly weak claim, but I wouldn’t go much further than that.
- Gartner asserts that, unlike software-only Oracle, Oracle Exadata isn’t significantly harder to administer than “other mixed OLTP/OLAP DBMS vendors,” because Exadata is fast enough you don’t need to jump through all those hoops any more to get tolerable performance. The money quote is “one reference reported reducing the number of indexes by a factor of 100 to fewer than five.” Note, however, that Gartner does not seem to assert that Exadata’s ease of use rivals that of the newer analytic DBMS specialists.
- Gartner confirms Oracle’s reluctance to do onsite Exadata POCs, but says it is not absolute. This is roughly compatible with what I’m hearing elsewhere, and indeed with Oracle own claims to be ramping up availability of Exadata POC hardware.
- Gartner’s criteria for inclusion include at least 10 different organizations having a product “in production.” Thus, the big surprise was ParAccel being included. The money quote there is “With approximately 20 customers in the pharmaceutical, retail, financial and media/advertising analytics sectors, ParAccel has a good reference base.” That assessment is difficult to reconcile with other information, but I’ve been told Gartner is sticking to its guns. That assessment would be even harder to believe if those 20 references were all alleged to be true production customers.
- Gartner notes that you basically can’t run a 1 TB+ MySQL data warehouse without sharding. (Of course, Infobright has an alternative, and up to a small number of terabytes so does Kickfire.)
- Gartner reports that at least some customers are pleased with Sybase IQ’s mixed workload/enterprise data warehouse capabilities.
- Gartner correctly notes that Oracle Exadata is a price-competition challenge for Teradata.
- Gartner notes that 20% of Vertica’s customers are outside the US. While not shocking, that’s more than I realized.
- Gartner notes something I don’t think I’ve posted yet, which is that Vertica has a customer with 300 TB of data. (The identity is a deep dark secret, but if I told you you probably wouldn’t recognize the name anyway.)
As does any such piece, the Gartner Data Warehouse DBMS Magic Quadrant also has outright errors. For example:
- Aster Data isn’t really “the newest entrant to the DBMS data warehouse world.”
- Aster’s SQL/MapReduce was not new in Release 4.0.
- Greenplum isn’t yet pushing down code to the storage tier.
- I’m not sure what kind of database-tier parallelism Gartner is claiming is new in Oracle in 11g Release 2 — but I doubt it’s really new. Rather, what Oracle has done recently is make parallelism less administratively cumbersome.
- Vertica wasn’t really the first DBMS in the cloud. At most it was the first pure-play analytic DBMS to get there.