Re: TPC-A and -B Benchmarks

In a recent append, obelix.informix.com!johnl@emory.mathcs.emory.edu (Jonathan Leffler) writes... >The whole point of setting the isolation level to repeatable read is to >command to the database system to ensure that any modification on any >table which triggers a modification on any of the data read by TX1 is >disallowed, because TX1 has demanded repeatable read isolation. If this >is not enforced, then what you have is not repeatable read as I >understand it. > >The issue rapidly reduces to: what does who mean by repeatable read. >Oracle supplies one definition; Informix supplies another; I can't And herein lies the crux of my question. It sounds to me (someone totally ignorant about Oracle 7) that Oracle 7 will have a hard time running the TPC-C benchmark because of its Repeatable Read implementation. Will it not have to obtain table locks on tables it modifies, to correctly implement the TPC-C benchmark? If this statement is true (and it very well might not be - I hope for corrections) then Oracle 7 would probably be at a great disadvantage when competing with Informix and others who do have Repeatable Read implemented with row locking, on the TPC-C benchmark. Their concurrency would be low. Pat Gainer gainer@almaden.ibm.com