Re: Number of extents related with performance?

"David Fitzjarrell" <oratune_at_aol.com> wrote in message news:92gg5k$3r2$1_at_nnrp1.deja.com...
[Snip]
>
> The old 'hard and fast' rules for extents are as follows:
>
> 2k block size 121 extents
> 4k block size 249 extents
> 8k block size 505 extents
>

Damn! You're right, David -clearly I'm going senile in my middle age! (though I *was* close!!)

> There is some logic to the extent maximums; starting with the 2k block
> size (the minimum allowed) the next block size will contain (2*old
> extents) + 7:
>
> 4k max = 2(121) + 7 = 242 + 7 = 249
> 8k max = 2(249) + 7 = 498 + 7 = 505
>
> Extrapolating to a 16K block size should provide a maximum of 1,017
> extents (a value I have not verified, although the previous values
> posted have been validated).
>

Yup, 1017 looks spookily familiar.

> This may not be a rigorous mathematical proof (it has been quite some
> time since I have dealt with Eigenvalues) but the calculation provides
> the correct number of expected extents given the 'power of 2' block
> sizes usually implemented.
>

Last time I dealt with Eigenvalues, I sent them packing... when last heard of, they'd taken up residence at the local bar, and were downing beers by the Stein. Dodgy lot, if you ask me.

Regards
HJR
> --
> David Fitzjarrell
> Oracle Certified DBA
>
>
> Sent via Deja.com
> http://www.deja.com/
Received on Thu Dec 28 2000 - 22:38:22 CST