Discrete Transactions and benchmark cheating

I can’t believe The Register is bringing up Discrete Transactions after all these years.  Yes, most of the industry thought "Discrete Transactions" were a benchmark special.  I was, I admit, incensed, about them.  The reason was simple, we couldn’t figure out any real application that could make use of them!  They were just so anti-Oracle in architecture, intent, and restrictions that the only reason we could see for their existence was to improve the benchmark score.  And that caused us (DEC) to push hard to add the benchmark special clause to the specification.

All products, including DEC (now Oracle) Rdb have features that were inspired by the TPC benchmarks.  And some of those have fairly narrow applicability in general applications.  But we used the test that their had to be some class(es) of real applications that could use the feature/tuning knob before we would add it to the product.  So we were in the spirit of "no benchmark specials" long before there was such a clause.  But that doesn’t mean we weren’t pushing the limit at every opportunity.  Just like rules in NASCAR and other sports, if something is open to interpretation you interpret it to your benefit.  At some point someone crosses the line and the rules are changed to reduce the size of the gray area.  That’s what happened with Discrete Transactions.  No big deal.

Ok, I can say "no big deal" now.  Back in 1993 it was a huge deal to me, DEC, and the entire database industry.  But why bring it up now?

This entry was posted in Computer and Internet. Bookmark the permalink.