Re: PGA_AGGREGATE_TARGET sizing of work areas and HJ cost
Date: Fri, 29 May 2009 21:55:41 +0200
joel garry pisze:
> On May 29, 10:45 am, "Jonathan Lewis" <jonat..._at_jlcomp.demon.co.uk>
>> >> Joel, >> >> From the way the OP asks the question I don't think he had the >> parameter hash_area_size in mind; to me it sounded like a >> question about "how is the equivalent of the hash_area_size >> derived when using automatic workarea sizing". >>
> From his reply to John, that's even less clear to me. I have no
> problem with myself or anyone misreading any of this, I think it is a
> good thing that people are willing to fill in the blanks. I read
> John's reply as asking if he's even using bind variables, it's only in
> the context of histograms that "using bind variables" becomes sensible
> as a plan changer. No offense intended towards Gregor, he needed a
> lot more info in the original post - histograms might make all the
> difference here, and his subsequent post references manipulating
> hash_area_size, perhaps as a way of telling John he knows about binds
> - or perhaps he is too quick to eliminate histograms as an issue. So
> I for one am confused about exactly what he is trying, and what level
> of expertise he has. You are probably less confused than me in
> general :-) but we need more details, replicable case, blabla, if
> anyone even really cares about 9.2. But of course, there may be
> something interesting here in more recent versions, too.
You're right , I have to paste some detailed info.
Its not bad so far I've learned that cost of HJ
is independent of PAT fluctuation. :D
and with sort_area_police = auto there is constant cost of HJ .
I'm out of work so left my problems too :D. Regards.
I'm regular reader Jonathan blog and few others. Received on Fri May 29 2009 - 14:55:41 CDT