Re: How are extents allocated in a multi-datafile tablespace

From: Michael Ray <Michael.Ray_at_trw.com>
Date: Thu, 21 Dec 2000 10:55:13 -0800
Message-Id: <10717.125162@fatcity.com>

--=_772C023B.20412D17

Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: quoted-printable
Content-Disposition: inline

I thought a similar thing (and have also heard or seen about the "auto-stri= ping"), and I don't really know the answer to your particular question =

offhand. I have a different twist on the subject. In my situation, I have =
autoextend on with a maxsize set. When my first datafile was at it's 15th =
of 16 possible extents, I created another similar datafile (though I just =

went for the whole 9 yards this time rather than let it autoextend). I was = expecting it to continue to fill up the first one and autoextend once more = and then use the second, but it never has.

So once my second fills up, I'm guessing it will finally autoextend the = first so it will use the remaining 250M. Either that or it will start = blowing errors about being out of space. 8o) I'm still pretty much a = newbie at this stuff so I'll just take it as it comes.

Michael Ray
Oracle DBA
TRW, Marshall, IL
217-826-3011 x2438

>>> chuck_hamilton_at_yahoo.com 12/21/00 11:17AM >>> It's always been my understanding that when a tablesapce has multiple = datafiles, extents are allocated in 1 datafile at a time until that = datafile is full, then it switches to the next datafile. Today I read a = note on a metalink forum where someone stated quite authoritatively that = extents are allocated in a striped fashion amongst the datafiles. For = example if you have three datafiles, the first extent goes to file 1, the = second to file 2, the third to file three, and then back to file 1. Is = this true? Is this something that's new in one of the versions of 8i? I've = never heard of such a thing. It would be great if that were true because = it would distribute i/o more evenly. But I thought the only way to = accomplish that in Oracle was with hash paritioning.

--=_772C023B.20412D17
Content-Type: text/plain
Content-Disposition: attachment; filename="TEXT.htm"

<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN">
<HTML><HEAD>
<META content="text/html; charset=iso-8859-1" http-equiv=Content-Type>
<META content="MSHTML 5.00.2919.6307" name=GENERATOR></HEAD>
<BODY style="FONT: 8pt MS Sans Serif; MARGIN-LEFT: 2px; MARGIN-TOP: 2px">
<DIV>I thought a similar thing (and have also heard or seen about the
"auto-striping"), and I don't really know the answer to your particular question offhand. I have a different twist on the subject. In my situation, I have autoextend on with a maxsize set. When my first datafile was at it's 15th of 16 possible extents, I created another similar datafile (though I just went for the whole 9 yards this time rather than let it autoextend). I was expecting it to continue to fill up the first one and autoextend once more and then use the second, but it never has.</DIV>
<DIV> </DIV>
<DIV>So once my second fills up, I'm guessing it will finally autoextend
the first so it will use the remaining 250M. Either that or it will start blowing errors about being out of space. 8o) I'm still pretty much a newbie at this stuff so I'll just take it as it comes.</DIV>
<DIV> </DIV>
<DIV>Michael Ray<BR>Oracle DBA<BR>TRW, Marshall, IL<BR>217-826-3011
x2438<BR><BR>>>> chuck_hamilton_at_yahoo.com 12/21/00 11:17AM >>><BR>It's always been my understanding that when a tablesapce has multiple datafiles, extents are allocated in 1 datafile at a time until that datafile is full, then it switches to the next datafile. Today I read a note on a metalink forum where someone stated quite authoritatively that extents are allocated in a striped fashion amongst the datafiles. For example if you have three datafiles, the first extent goes to file 1, the second to file 2, the third to file three, and then back to file 1. Is this true? Is this something that's new in one of the versions of 8i? I've never heard of such a thing. It would be great if that were true because it would distribute i/o more evenly. But I thought the only way to accomplish that in Oracle was with hash Received on Thu Dec 21 2000 - 12:55:13 CST