There is a limit (undocumented, but mentioned in the past on this list) of 8K bytes for any expression in DataStage. If its getting near this limit (say over 6K bytes, since column names are re-mapped when the expression is compiled), consider replacing the expression with a call to a Routine (transform function).
In the "Advanced DataStage" class offered in Asia-Pacific region there is a module on the topic "Designing for Performance". This is the only resource of which I am aware.
The main performance issues typically are external factors, such as available (rather than theoretical) network bandwidth, database server performance and hardware performance. Of course, inefficient coding practices in DataStage will never help!
-----Original Message-----
From: Alexandre Stoulov [mailto:alexs@tern.ru]
Sent: Monday, 02 April 2001 18:24
To: informix-datastage@oliver.com
Subject: Some questions
Hi, all!
First of all I need your advise about any resources concerning DS perfomance tuning (books, url, etc)
The second question is more technical: Is there any limitations on the output constraints in transform stage. We find unpredictable behaviour of the job if the length of constraint is large enough.
Thanks in advance, Alexandre Stoulov
------------------------------
TERN Technical Support
mailto:alexs@tern.ru
Some DataStage questions
Some DataStage questions
Try the Compare() function, rather than a comparison operator. It may also be that there are non-printing characters in one of the strings, as suggested earlier. Use Oconv(string,MCP) to convert any non-printing character to ".", to see whether this is so. However, none of this seems related to the poor performance you are seeing. There are no tools specifically for monitoring DataStage; it is an application on the machine on which it is installed. You can monitor this with conventional tools, such as sar. On Windows NT there is the performance monitor, and there are hooks in DataStage (UniVerse) that can be used to count events into the performance monitor, but these can not be introducted directly into an expression.
-----Original Message-----
From: Alexandre Stoulov [mailto:alexs@tern.ru]
Sent: Monday, 02 April 2001 19:25
To: Ray Wurlod
Subject: Re[2]: Some DataStage questions
Thanks, Ray!
But this is very strange. The length of constraints is about 1K. The condition is simle, it check the equality of string in DB and inut string and works wrong (debuger shows identical values, Ardent considers conversly).
I dontt think that in our case (Sun Enterprise 3500, 2Gb Ram,
IDS.2000) performance is determined by external factors. The simple row update in the 50000 records table longs more then 1 second!!. While queries produced by dbaccess, Business Objects are executed with normal speed.
May be you know any tools, methods, etc to monitor Ardent perfomance and find bottleneck?
Alexandre
RW> There is a limit (undocumented, but mentioned in the past on this
RW> list)
of
RW> 8K bytes for any expression in DataStage.
RW> If its getting near this limit (say over 6K bytes, since column
RW> names
are
RW> re-mapped when the expression is compiled), consider replacing the
RW> expression with a call to a Routine (transform function).
RW> In the "Advanced DataStage" class offered in Asia-Pacific region
RW> there
is a
RW> module on the topic "Designing for Performance". This is the only
resource
RW> of which I am aware.
RW> The main performance issues typically are external factors, such as
RW> available (rather than theoretical) network bandwidth, database
RW> server performance and hardware performance. Of course, inefficient
RW> coding practices in DataStage will never help!
RW> -----Original Message-----
RW> From: Alexandre Stoulov [mailto:alexs@tern.ru]
RW> Sent: Monday, 02 April 2001 18:24
RW> To: informix-datastage@oliver.com
RW> Subject: Some questions
Best regards,Alexandre Stoulov
------------------------------
TERN Technical Support
mailto:alexs@tern.ru
-----Original Message-----
From: Alexandre Stoulov [mailto:alexs@tern.ru]
Sent: Monday, 02 April 2001 19:25
To: Ray Wurlod
Subject: Re[2]: Some DataStage questions
Thanks, Ray!
But this is very strange. The length of constraints is about 1K. The condition is simle, it check the equality of string in DB and inut string and works wrong (debuger shows identical values, Ardent considers conversly).
I dontt think that in our case (Sun Enterprise 3500, 2Gb Ram,
IDS.2000) performance is determined by external factors. The simple row update in the 50000 records table longs more then 1 second!!. While queries produced by dbaccess, Business Objects are executed with normal speed.
May be you know any tools, methods, etc to monitor Ardent perfomance and find bottleneck?
Alexandre
RW> There is a limit (undocumented, but mentioned in the past on this
RW> list)
of
RW> 8K bytes for any expression in DataStage.
RW> If its getting near this limit (say over 6K bytes, since column
RW> names
are
RW> re-mapped when the expression is compiled), consider replacing the
RW> expression with a call to a Routine (transform function).
RW> In the "Advanced DataStage" class offered in Asia-Pacific region
RW> there
is a
RW> module on the topic "Designing for Performance". This is the only
resource
RW> of which I am aware.
RW> The main performance issues typically are external factors, such as
RW> available (rather than theoretical) network bandwidth, database
RW> server performance and hardware performance. Of course, inefficient
RW> coding practices in DataStage will never help!
RW> -----Original Message-----
RW> From: Alexandre Stoulov [mailto:alexs@tern.ru]
RW> Sent: Monday, 02 April 2001 18:24
RW> To: informix-datastage@oliver.com
RW> Subject: Some questions
Best regards,Alexandre Stoulov
------------------------------
TERN Technical Support
mailto:alexs@tern.ru