Re: Informix beats Oracle

Obnoxio The Clown wrote: > DA Morgan said: >> How about a simple test ... inserting records into a table? Could it be >> simpler? > > [SNIP] > >> Wow, look at that Oracle 10g is 3000 times faster than Oracle 10g. >> Maybe I should switch from one to the other. >> >> Benchmarks. ROFLOL! > > And just for the sake of the argument, Daniel, what if I ran a side by > side comparison of Oracle 10g and IDS 10.0 using exactly the same C source > code, with only the Pro*C headers and error functions replaced by their > ESCL/C equivalents? On exactly the same tin, not even comparably > configured but different servers? > > Would that constitute a fair benchmark? It would be meaningless because "best practice" in Oracle would be using techniques not available in Informix and I would expect "best practice" in Informix would similarly use techniques not used in Oracle. You must remember that the concepts and architecture of the two have been quite different. What might constitute a fair test is "best practice" in both products with Cheetah's new isolation level that comes close to functionally matching Oracle's MVCC. But I'd want to put the production release into my lab and verify that they are equivalent or close to it before stating that with confidence. The issue you may notice above is "best practice." Most Oracle DBAs and developers, unfortunately, cut their teeth on version 7 and still code and optimize using techniques that are "backward compatible" ... not "best practice." Likely the same holds true in the Informix world as well. The issue, if you want to benchmark, is not does it compile but is it the state of the art for the product: Is it what the product is capable of doing. But in all of the focus on speed I wonder whether you are ignoring the large number of other comparisons that are equally critical such as the ability to seamlessly enrypt data, encrypt backups, and meet other legal requirements around auditing and compliance. There is more to sitting under an RDBMS than just DML. -- Daniel A. Morgan University of Washington damorgan@x.washington.edu (replace x with u to respond)

DA Morgan said: > Obnoxio The Clown wrote: >> DA Morgan said: >>> How about a simple test ... inserting records into a table? Could it be >>> simpler? >> >> [SNIP] >> >>> Wow, look at that Oracle 10g is 3000 times faster than Oracle 10g. >>> Maybe I should switch from one to the other. >>> >>> Benchmarks. ROFLOL! >> >> And just for the sake of the argument, Daniel, what if I ran a side by >> side comparison of Oracle 10g and IDS 10.0 using exactly the same C >> source >> code, with only the Pro*C headers and error functions replaced by their >> ESCL/C equivalents? On exactly the same tin, not even comparably >> configured but different servers? >> >> Would that constitute a fair benchmark? > > It would be meaningless because "best practice" in Oracle would be using > techniques not available in Informix and I would expect "best practice" > in Informix would similarly use techniques not used in Oracle. You must > remember that the concepts and architecture of the two have been quite > different. It's got nothing to do with "best practise", Daniel. The application is the application. It's simple, it's written in C, there are no stored procedures, there are no tricks, no cleverness. It's the sort of thing someone would write as departmental or small- to mid-range business application. It's just load on the database, sticking rows in, getting rows out, occasionally updating them or deleting them. There is no "best practise" screwing around allowed. > What might constitute a fair test is "best practice" in both products > with Cheetah's new isolation level that comes close to functionally > matching Oracle's MVCC. But I'd want to put the production release into > my lab and verify that they are equivalent or close to it before stating > that with confidence. It's entirely irrelevant Daniel. Even with Oracle's "superior" MVCC to IDS 10's crappy old shit, and even with checkpoints factored in, it was still better, faster and cheaper. > The issue you may notice above is "best practice." Most Oracle DBAs and > developers, unfortunately, cut their teeth on version 7 and still code > and optimize using techniques that are "backward compatible" ... not > "best practice." Likely the same holds true in the Informix world as > well. The issue, if you want to benchmark, is not does it compile but is > it the state of the art for the product: Is it what the product is > capable of doing. Bullshit Daniel. The only valid benchmark is the application you are running or are going to run. If that application meets your business criteria then it doesn't matter whether it's capable of using the database to the best of the database's capabilities. People will never buy SAP because it's capable of using every aspect of DB2. The application is the application. If it's such a simple application that even I can understand it, then it's very simple indeed. You can't hide behind "superior Oracle technology that hasn't been used". > But in all of the focus on speed I wonder whether you are ignoring the > large number of other comparisons that are equally critical such as the > ability to seamlessly enrypt data, encrypt backups, and meet other legal > requirements around auditing and compliance. There is more to sitting > under an RDBMS than just DML. You're bringing up irrelevancies, Daniel. The application is the application. -- Bye now, Obnoxio "I'm astonished anyone pays real money for this crap." -- Cosmo -- This message has been scanned for viruses and dangerous content by OpenProtect(http://www.openprotect.com), and is believed to be clean.