February 18, 2010 at 1:22 am
Hi, we have run into a performance issue with a master package calling child packages and would really appreciate some advice.
We have three SSIS projects, each consisting of a set of child packages and a master package that performs some admin tasks and then call all the child packages. All of the child packages are executed with ExecuteOutOfProcess=false.
Master1: admin task -> execute package1.1 -> execute package1.2 …-> execute package1.n-> admin task;
Master2: admin task -> execute package2.1 -> execute package2.2 …-> execute package2.m-> admin task;
Master3: admin task -> execute package3.1 -> execute package3.2 …-> execute package3.k-> admin task;
When executed separately using the Execute Package Utility, the execution times for these packages are (mm:ss)
Master1 05:27
Master2 25:42
Master3 11:50
We want to schedule these packages and add some additional conditions, however, simply creating a new Master package that calls these 3 packages sequentially (obviously using the same data sets), increases the execution times to
Master1 05:29 = an increase of 0.5%
Master2 32:33 = an increase of 27 %
Master3 36:28 = an increase of 208% !!.
We have tried to set ExecuteOutOfProcess=true on the Execute Package tasks but that only increased the execution time further.
I’m not sure what causes this and was not able to find anything similar on the forums.
Any help would be appreciated.
Thanks
February 18, 2010 at 1:33 am
It would be useful background if you would explain a bit about what the packages are doing - what is Master3 doing that the other 2 are not?
The absence of evidence is not evidence of absence
- Martin Rees
The absence of consumable DDL, sample data and desired results is, however, evidence of the absence of my response
- Phil Parkin
February 18, 2010 at 1:38 am
The packages all perform the same type of tasks - reading from database tables, performing transformations and loading data into tables. There are no significant differences.
February 18, 2010 at 2:01 am
Then there is a suggestion that the time increase gets worse as each package runs.
Can you swap, say, Master1 and Master3 to confirm this?
It would be useful to know whether it's a general degradation, as the numbers suggest, or a package-specific degradation.
The absence of evidence is not evidence of absence
- Martin Rees
The absence of consumable DDL, sample data and desired results is, however, evidence of the absence of my response
- Phil Parkin
February 18, 2010 at 7:13 am
OK, so I have set up a test package that calls master1 4 times in a row while resetting things like target database tables to their original state in between calls. These are the results:
Master1 call1: 04:40
Master1 call2: 05:42 = increase of 22%
Master1 call3: 06:44 = increase of 44%
Master1 call4: 08:29 = increase of 82%
There has to be some sort of resource problem
February 18, 2010 at 9:24 am
[font="Comic Sans MS"]
Yes ... I am using 3 level of package hierarchy and performancewise it has been absolutely fine...
[/font]
[font="Comic Sans MS"]--
Sabya[/font]
Viewing 6 posts - 1 through 5 (of 5 total)
You must be logged in to reply to this topic. Login to reply