Re: Estimate or analyze statistics

Probably not. That's the reason ESTIMATE is provided in the first place. Although you could try the COMPUTE option if you have a powerful and plentiful hardware and a big enough time window (which, of course, almost never happens). When you ESTIMATE, 25% is a good minimum, if possible. However, I have a table that's approaching 200 Gig, and I am estimating 10%. In my opinion, effective data warehousing (on enterprise scale) can be achieved only with sufficient hardware. For large corporate DWs, a production system is almost always MPP. Anyways, to be able to COMPUTE every time, you may need a monster system with lots of available horsepower. I sure wish I did. :)

Having said that, if you can partition your tables and the new data is always appended and the old data is never changed then you can COMPUTE only for the last partition created (if, again, it's small enough).

To figure out what's really feasible in your case, try estimating some small percentage and see how long it takes. That should give you an idea of what 100% might take.

HTH
Sergey

In article <3A00F67C.719DE5C0_at_yahoo.com>,   hellothere <hellothere_at_yahoo.com> wrote:
> Running Oracle 8.0.5 on Solaris 2.6
> I apologize in advance if this has been addressed, but are there any
> hard and fast rules
> about estimating vs. analyzing statistics? I've seen table size
> mentioned, as well data skew.
> Is it even feasible to analyze and compute statistics on tables in the
> multiple gigabyte range (20+ GB)?
>
> TIA
>
>

Sent via Deja.com http://www.deja.com/
Before you buy. Received on Thu Nov 02 2000 - 06:57:30 CST