Re: IBM releases #1 TPC-E benchmark!

On Jul 15, 12:40 am, PeterP <peterp...@gmail.com> wrote: > On Jul 14, 11:22 pm, Ian Michael Gumby <im_gu...@hotmail.com> wrote: > > > > > On Jul 14, 5:48 am, Serge Rielau <srie...@ca.ibm.com> wrote:> DA Morgan wrote: > > > >> Your conspiracy theories are getting really annoying. > > > > Annoyingly accurate methinks. > > > > > Any time you don't like it feel free to meet the demands of the user > > > > community. Some people call them "customers." > > > > I was wondering when you'd show up. > > > Although I expected you to help discuss temp tables ;-) > > > Poor Serge. > > > I mean yeah, I could pick on Daniel, but really Mark is the one who's > > in a position to fix Oracle. > > I mean heck, they eventually got their R-Tree issue solved in 10g. So > > one day they'll get their temp tables figured out. The only thing is > > that I wonder who will fix their temp tables first. DB2 or Oracle? ;-) > > > But lets get back on track. > > > How can IBM continue to press flesh in Lenexa, sing the praises of the > > IIUG and IDS. Yet, well, no spend real money on marketing. Not do the > > benchmarks or rather release the benchmarks. And well frankly put in > > place DB2 specialized teams to promote the brand. > > > And here's poor Serge. I just wonder what he was doing when DB2 V8 was > > released on distributed? Or more to the point, what he was working on > > when Janet tried to convince a customer to switch from XPS to DB2 V8. > > > I know a couple of stories that would make Serge blush. BTW what's Bob > > P doing these days? Didn't he take the heat for that blunder?  And why > > was Janet in such a rush to release "Viper". (DB2 8 was Viper, right?) > > Some actual figures one day would be good, from a personal point to > see aah, that’s what numbers can be on this setup. Managers / decision > makers often think differently though. > Perhaps it's because IDS is much faster / cheaper TCO than DB2 > (although steeper IDS prices now favour DB2) and IBM doesn't want to > annoy all their people on DB2 to have them switch to SQL (well SQL > would be supported and not expensive with constant barrage marketing > like Oracle) > The other thing is, if IBM decided to do an IDS benchmark, an exec > type at IBM is going to look at TPC and say, right top score is 6 > million, 10 million sounds good to me this morning. IDS will have to > go through somewhat of a re-write (rather than useful code > development) and the drivers on the hardware it's running having > similar tweaks until the magic burn up chart hits the figure. I'd be > frowning on this development approach, wondering is this good for a > product? Or are benchmarks actually done some other way - i.e. > actually testing a bog standard product on a standard non optimised > hardware setup. :) > On second thoughts, if it means more mainframe dollars for IBM the > exec would end up saying... right DB2 is 10 mil and in the lead, IDS > was better TCO / quicker but post a benchmark showing it less than DB2 > so not to affect our bottom line which is no use for Informix either. > It would be nice to do strings oninit and find an ids_db2_tco_ratio > undocumented variable - yeah I'm sure it would be in the makefile / > typedef. > The more you think about benchmarks, the more it seems a waste of time > - damn, just bothered to write this. Actually, this variable already exists... in db2: inefficient_page_structure = 1 or in ids: page_compression = 0 or is it multi threading that's the biggest gain of recent in 9.5?