RE: Export direct=Y|N

I don't know the answer for sure, but can hazard a guess. :-)

65535 is typically the maximum number of bytes that can be written with a single write() system call. If you use a shorter record length, you'll execute multiple write()'s for each 64K of data, which involves multiple trips through kernel land. If you use larger values, you're not buying anything unless it's a multiple of 65535; i.e. if you specify 65536 you'll be back to 2 writes per chunk. Specifying 2x65535 will require the same two writes, and gets 65534 more bytes transferred than the previous example!

It's a very small incremental time, but over the course of a 20G export all those round trips between kernel and user space can really add up....

--
Rich Holland        (913) 645-1950        SAP Technical Consultant
print unpack("u","92G5S\=\"!A;F]T:&5R(\'!E<FP\@:&%C:V5R\"\@\`\`"); 

-----Original Message-----
From: root_at_fatcity.com [mailto:root_at_fatcity.com] On Behalf Of Koivu,
Lisa
Sent: Friday, May 23, 2003 10:25 AM
To: Multiple recipients of list ORACLE-L
Subject: RE: Export direct=Y|N


Steve, 

Why?

Lisa
-----Original Message-----
From: Steve Perry [mailto:sperry_at_sprynet.com]
Sent: Thursday, May 22, 2003 9:17 PM
To: Multiple recipients of list ORACLE-L
Subject: Re: Export direct=Y|N


I'm still catching up so somebody may have added to this, but adding
recordlength=65535 with the direct=y will improve it even more.

----- Original Message ----- 
From: Wiegand, Kurt 
To: Multiple recipients of list ORACLE-L 
Sent: Thursday, May 22, 2003 8:26 AM
Subject: Export direct=Y|N


I ran across an export that was taking 9 1/2 to 10 hours to complete
(the resulting dmp 
file was about 7GB compressed).  I added DIRECT=Y and it now takes 3 1/2
hours.  My 
question is: What are the reasons that someone would choose not to use
DIRECT=Y? 
This is an 8.1.5.0.0 database running on a Solaris box; I would hate to
have just removed a 
work around for some 8.1.5.0.0 bug!  I have found comments in some
scripts here, but 
they all seem to be in my handwriting!! 
Kurt      kurt.wiegand_at_cwusa.com 


A ran a diff on the log files after the change.  Here is the beginning
(just what I expected)..... 
23c23,24 
< . about to export SYSTEM's tables via Conventional Path ... 
--- 


> . about to export SYSTEM's tables via Direct Path ... 

> EXP-00067: Table DEF$_AQCALL will be exported in conventional path. 

24a26 


> EXP-00067: Table DEF$_AQERROR will be exported in conventional path. 

29a32 


> EXP-00067: Table DEF$_LOB will be exported in conventional path. 

33a37 


> EXP-00067: Table DEF$_TEMP$LOB will be exported in conventional path. 

127c131 
< . about to export OUTLN's tables via Conventional Path ... 
--- 


> . about to export OUTLN's tables via Direct Path ... 

130,136c134,140 
< . about to export DBSNMP's tables via Conventional Path ... 
< . about to export OPS$ORACLE's tables via Conventional Path ... 
< . about to export CW30_AUDITOR97's tables via Conventional Path ... 
< . about to export CW30_LAC_MANUAL's tables via Conventional Path ... 
< . about to export DISPUTES2's tables via Conventional Path ... 
< . about to export DISPUTES3's tables via Conventional Path ... 
< . about to export BANK's tables via Conventional Path ... 
--- 


> . about to export DBSNMP's tables via Direct Path ... 

> . about to export OPS$ORACLE's tables via Direct Path ... 

> . about to export CW30_AUDITOR97's tables via Direct Path ... 

> . about to export CW30_LAC_MANUAL's tables via Direct Path ... 

> . about to export DISPUTES2's tables via Direct Path ... 

> . about to export DISPUTES3's tables via Direct Path ...