The T
ransRelational(TM) Model is an approach to implementing the
RelationalModel developed by Stephen Tarin, based on his patented Tarin Transform Method. It is claimed to offer improved physical data independence and performance over various popular strategies for implementing the
RelationalModel, particularly when joining relations. An overview of how it works can be found in Appendix A of the 8th edition of
AnIntroductionToDatabaseSystems, where
ChrisDate claims that it "is likely to prove the most significant development in this field since
TedCodd gave us the relational model..." It is described in considerably more detail in
ChrisDate's online book on the T
ransRelational(TM) Model at
http://bookboon.com/en/textbooks/it-programming/go-faster
However, at the time of this writing there are apparently no available DBMS products based on it. Allegedly, a prototype or pre-release product was developed by Tarin's company, Required Technologies Inc. (
http://www.requiredtech.com/), but nothing was commercially released -- apparently for unspecified business reasons not related to the approach itself.
Furthermore, there appear to be no published academic analyses or experimental verifications of its performance. At least one preliminary analysis (
http://www.dbforumz.com/TransRelational-Model-Performance-Concerns-ftopict169586.html) suggests that actual implementations may not meet the performance claims. This has resulted in an ongoing debate between
FabianPascal, Curt Monash, and others over the value of the approach, the accuracy of the claims made about it, its applicability to general purpose relational DBMSes as opposed to write-once read-many analytical tools, and the nature of the aforementioned prototype.
The name "T
ransRelational" was apparently chosen for marketing reasons and intended to be evocative of
StarTrek's "T
ransWarp Drive", though "Trans-" is apparently intended to represent "transform".
See:
I thought I saw it mentioned on Slashdot a month or so ago. Anyhow, I suspect it won't go anywhere until a
BigIron RDBMS company buys them out and takes the time necessary to nurture it.
If it provided a fairly significant gain, you'd think the middle or BigIron DB's co's would have already snapped up the company, or at least licensed the technology. It would give them an advantage over competitors. Thus, I suspect it's not living up to the hype. Bumbling infighting wouldn't matter much if the tech was good because DB co's would just buy the company and fire or silence the problem managers.
It appears to me that it merely offers performance trade-offs rather than an over-all net gain. There may be niches where it performs significantly better, but are these niches common enough to justify creation and maintenance of a second engine that is query-language-compatible with a traditional implementation offered by a vendor? (I don't think one would want to bet the whole company on it.) And is this more economical than throwing bigger hardware at a problem? High-end existing RDBMS offer lots of options for tuning-to-task such that a skillful tuner may be able to get it close to TRM performance.
There's now a big
KleinerPerkins-funded startup called
VerticaSystems whose forthcoming
VerticaDatabase (
http://www.vertica.com/) may have some connection to
TransRelational. It's apparently going to be column-oriented, and the marquee technical name on the project is
MichaelStonebraker, who was apparently involved with
TransRelational too (see
http://en.wikipedia.org/wiki/TransRelational_model#Commercial_implementation). On the other hand
VerticaSystems is claiming patents that date back to 1991, and
StephenTarin doesn't seem to be involved, so the technical and business relationship may be murky.
Apparently, the Vertica DBMS has no connection to the TransRelationalModel, being based on this: http://db.lcs.mit.edu/projects/cstore/vldb.pdf
CategoryDatabase CategoryModels