RE: Index rebuilding
- From: "Cary Millsap" <cary.millsap@xxxxxxxxxx>
- To: <oracle-l@xxxxxxxxxxxxx>
- Date: Sun, 14 Nov 2004 10:46:21 -0600
Doesn't it come down to making sure you've defined your terms? A lot of =
the
argument seems to be an implicit disagreement over what the word =
"balanced"
means. In Knuth and other computer science texts that discuss indexes, I
believe the definition of "balanced" is "an index is balanced iff (if =
and
only if) all leaf nodes have the same distance to the root." By this
definition, Oracle B*-tree indexes are ALWAYS balanced, and NEVER
un-balanced. This point is not in contention, correct?
I think what's happening is that people who are complaining about
un-balanced-ness are redefining the word "balance" to mean something
completely different.
In general, I think it's sloppy to take change the meaning of a =
scientific
word in a discussion or "white paper." When I say "scientific word," I =
mean
one that has been carefully defined and used in a specific context =
for--in
this case--decades. It's one of the things that drives me nuts about the
Oracle culture, this bastardization of carefully defined, =
well-established
terms for the convenience of some Oracle author who writes more than he
reads. :)
I guess the problem is analogous to the one being solved in the XML =
world by
the implementation of XML namespaces. Maybe instead of the term =
"balanced",
we should use the term "knuth:balanced" or "choose-an-author:balanced". =
In
this case, I would suggest that the default namespace should be set to
"knuth".
Cary Millsap
Hotsos Enterprises, Ltd.
http://www.hotsos.com
* Nullius in verba *
Upcoming events:
- Performance Diagnosis 101: 1/4 Calgary, 2/2 Sydney
- SQL Optimization 101: 11/8 Dallas, 12/13 Atlanta, 2/7 Sydney
- Hotsos Symposium 2005: March 6-10 Dallas
- Visit www.hotsos.com for schedule details...
-----Original Message-----
From: oracle-l-bounce@xxxxxxxxxxxxx =
[mailto:oracle-l-bounce@xxxxxxxxxxxxx]
On Behalf Of Alex
Sent: Friday, November 12, 2004 5:44 PM
To: DGoulet@xxxxxxxx; oraclel@xxxxxxxxxx; oracle-l@xxxxxxxxxxxxx
Subject: RE: Index rebuilding
I agree with Dick! Always and never are to be used in
cases like "the sun always rises in the east: or "I've
never enjoyed working with Oracle more than I do now"
:)
Regards!
--- "Goulet, Dick" <DGoulet@xxxxxxxx> wrote:
> Looked at Richard Foote's paper. Don't know about
> that. I did prove to
> OTS several years ago that a block could get "lost"
> in an index due to
> deletion/updates that left it empty. I believe that
> got finally fixed
> in Oracle 8i. I've still seen cases of index's
> becoming unbalanced, I
> know the docs day it's impossible, but it does
> happen without the index
> height increasing. And I still believe that index
> deletes don't get
> flushed so efficiently, as Richard suggests. If
> that was the case then
> I can't explain why an index rebuild can cause an
> index to shrink by 30%
> or more. And recent experience still shows that a
> rebuild can cause
> significant performance improvement. And Oracle has
> provided the
> capability to rebuild indexes which is not trivial.=20
> Therefore, NEVER
> use the word "never" unless your absolutely certain
> that under all
> circumstances it will be absolutely true. And in
> the current context,
> that is the truth, that is, never can never be an
> absolute.
>=20
> BTW: Since we've a few "myth busters" in the group.=20
> I appreciate the
> effort these people put into "myth busting", even if
> they are later
> proven to have erred. At a very minimum they start
> discussion and
> re-examination of commonly held beliefs that can
> have changed or lost
> significance over the years(like it's best to have
> all of a tables data
> in the first extent). Such discussion, although
> sometimes the start of
> "Holy Wars", is healthy (not the Holy War though)
> and a necessary part
> of all of us growing. That being said, let it be
> noted that I agree to
> disagree, in part, with Mr Foote.
>=20
>=20
> Dick Goulet
> Senior Oracle DBA
> Oracle Certified 8i DBA
> -----Original Message-----
> From: Jared Still [mailto:jkstill@xxxxxxxxx]=3D20
> Sent: Friday, November 12, 2004 12:44 PM
> To: oraclel@xxxxxxxxxx
> Cc: oracle-l@xxxxxxxxxxxxx; steve@xxxxxxxxxxxxxx
> Subject: Re: Index rebuilding
>=20
> On Fri, 12 Nov 2004 11:49:46 +0100, Karsten Weikop
> <oraclel@xxxxxxxxxx>
> wrote:
> > Please read the execellent paper from Richard
> Foote (which can be
> > downloaded from Miracle's site):
> >
>
http://www.miracleas.dk/images/upload/Docs/Richard%20Foote.pdf
> > Conclusion form this paper: Never Rebuild, but
> find the course to the
> > problem.
>=20
> Never?
>=20
> I think you will find that statement as difficult to
> support as
> 'always rebuild'.
>=20
> --=3D20
> Jared Still
> Certifiable Oracle DBA and Part Time Perl Evangelist
> --
> //www.freelists.org/webpage/oracle-l
> --
> //www.freelists.org/webpage/oracle-l
>=20
=09
__________________________________=20
Do you Yahoo!?=20
Check out the new Yahoo! Front Page.=20
www.yahoo.com=20
=20
--
//www.freelists.org/webpage/oracle-l
--
//www.freelists.org/webpage/oracle-l
Other related posts:
