Teradata decides to compete head-on as a data warehouse appliance vendor
In a press release today that is surely timed to impinge on the Netezza user conference news cycle, Teradata has come out swinging. Highlights include:
- Teradata, which long avoided the “appliance” term, now says it sells both “data warehouse appliances” and “data mart appliances.” Indeed, it claims to have “invented the original appliance” — which is pretty close to being true.*
- Teradata claims its “new appliance easily delivers up to 5 to 10 times performance improvement over competitors’ appliances,” at $119,000 per terabyte US list price.
- Teradata claims a 150% faster “scan rate” than competitors. Teradata is surely thinking of Netezza when saying that.
- Teradata claims 10X performance improvement on “selected queries” vs. the “competition.”
- Teradata thinks its geospatial data management capability is better than competitors’, and that this is an important indicator of Teradata’s general overall greater sophistication.
*Pay no attention to Teradata’s unfortunate choice elsewhere in the release to say it’s introducing a “second-generation” appliance, and seemingly implying it offered a “first-generation” one in April.
I talked with Teradata’s chief development officer Scott Gnau Friday to clarify those performance claims. Here’s what Teradata means by them.
- Scott says a Teradata appliance really can do sequential reads 50% faster than other vendors, assuming the same disk drives. He gave an NDA/trade secret reason as to how Teradata does this that sounded plausible. However, I didn’t grill him as to whether this claim holds up in the face of different block-size choices in the file systems or different compression numbers. (Note: More compression might mean more ability to confine data to the faster outer bands of the disk.)
- “150% faster” on the scans really does mean 2.5X, at least when referring to those competitors that use slower disks than Teradata does.
- Another major component of the 5-10X claim is multidimensional partitioning. Frankly, I suspect Scott of underrating his competitors’ products in that regard, and anyhow the effect of partitioning is highly dependent on workload. (But hey, that’s what POCs are for!)
- The 10X claim is for certain queries in certain workloads as compared with certain other systems. (The number could obviously have been a lot higher if Teradata wanted to phrase things as ridiculously as some other vendors do.)
Please also note that vendors are always imprecise when they give numbers like these, and Teradata is no exception to that rule.
Comments
4 Responses to “Teradata decides to compete head-on as a data warehouse appliance vendor”
Leave a Reply
Another Appliance with Geospatial Support…
Just a day after Netezza announced its spatial extension for its appliance, Teradata (one “r”) notes (press release) it too has such support: “The Teradata 12 software foundation provides full features and advanced functionality for traditional anal…
[…] Last week, Teradata claimed it has the most sophisticated analytic geospatial data management capability. […]
[…] На прошлой неделе (статья написана 26.09.2008 – прим. ред.), Teradata заявила, что у неё самые большие возможности для современноой анал…. […]
[…] DV and BI users will need a lot of DWA for their “big data”. Teradata claimed 2 years ago that Netezza are far behind performance-wise, but apparently IBM disagrees or does not […]