11-25-09, 12:47 #1Registered User
- Join Date
- Feb 2009
Unanswered: named pipe to sqlldr reducing block size
I am spawning sqlldr from our windows application to load a fairly large table. When loading from a delimited file, I can specify the row/buffer sizes such that the commit occurs every thousand records or so. However, when I tell sqlldr to read from a named pipe, it commits much more frequently and performance is very low. This is surprising because of the many references in forums and documentation to using named pipes as a performance improvement. Apparently, when reading from a pipe on Windows, sqlldr only gets "pipesize" bytes at a time, and rather than reading again until its buffer is full, it decides to commit what it has. Is there a way to make this perform better?
11-25-09, 12:52 #2Registered User
Provided Answers: 1
- Join Date
- Aug 2003
- Where the Surf Meets the Turf @Del Mar, CA
>Is there a way to make this perform better?
use LinuxYou can lead some folks to knowledge, but you can not make them think.
The average person thinks he's above average!
For most folks, they don't know, what they don't know.
Good judgement comes from experience. Experience comes from bad judgement.