Re: RDBMS Performance figures. - please post replies

> > Well - title says it all. > > /Jacob > > Good luck. Consider, each RDBMS is going to perform differently according to what platform it is run on, the qualifications of the person settingup the environment for the test, the amount of data, how the platform is configured, yadiyadiya. The closest you can get to true numbers are the TPC benchmarks. These present a number of different measures, prime amongst these are the transactions per second and cost per transaction. There are three established benchmarks in the RDBMS world, TPC-A, B, and C. The first of these (A) is fairly old at this point. I forget exactly what it measured - I'm sure someone will jump in to help me here - essentially it was a measure of raw power for a single threaded process against the engine. (B) was a measure which included multiple users transacting against the engine. It also is getting long in the tooth. Both A and B are now considered obsolete. (from a press release not too long ago) > TPC-C Benchmark Predicts Real-World Database Performance > > The TPC-C benchmark, developed in 1992 by the Transaction Processing > Performance Council (TPC), is considered to be the best benchmark > test for evaluating real-world application performance. The benchmark > measures a more complex, transaction-intensive, business environment > by simulating an order-entry application in an OLTP environment. > The benchmark includes full OLTP emulation from on-line users to > storage devices; a mix of read-only and update intensive transactions; > and tests five transaction types including on-line and batch. The > benchmark results report the number of orders processed each minute > within a specified response time. For a time I saved off all of the press releases that floated across this group on Informix performance. Checking my save directory I see that I no longer do this. Ok, so I've convinced you to look at TPC-C (and 'D' I gather - though I have yet to see a TPC-D result). You still have to consider how each vendor presents themselves. They can try to get the highest trsanactions/second or go for the lowest cost/transaction. Every press release that you find from either a vendor or from Informix is going to say how so and so has now presented the best results ever for such and such benchmark - you're going to have to read these carefully to extract what you want. I'm not saying that this is a bad thing. What do I want 9 million transactions per second for when I'm doing a movie rental system? What I'm trying to say is that you should not pay too much attention to raw performance figures. In the long run there will be little to differentiate one HW/RDBMS combination from some of the others. What you should be looking at is who you want to be there to provide support to what you've got. Who understands the products in question better than the others? This will vary according to locale. Go ahead and ask folks on this group how they feel about supporting any RDBMS when their vendor can't cut the mustard. Some time ago, in a location which shall remain unamed, one of my clients bought a decent box with an Informix product and a version of Unix. They had a decent setup, but the HW vendor didn't a) know how to spell Unix, b) didn't know how to spell Informix. As a result they never got the support they required except through a third party (me) and they found me too expensive to handle day to day problems. In the end they wound up tossing the system - which was perfectly capable of handling their expanded needs given some upgrading - and went with a vendor who understood what they needed out of life. They are still quite happy with theit vendor and the system. The system is getting upgraded as it needs it, the RDBMS is working well with the HW and so forth. So while performance figures are definitely nothing to be sneered at, you should keep your eye on the overall picture. If you really want TPC figures you might try asking William Daul. cheers j. _____________________________________________________________________________ Jack Parker - Hewlett Packard, DMD/IS Boise, Idaho, USA jparker@hpbs3645.boi.hp.com _____________________________________________________________________________ Back up my hard drive? You mean these things can actually run in reverse? _____________________________________________________________________________ Any opinions expressed herein are my own and not those of my employers. _____________________________________________________________________________