SAP BI 7 - Background process performance Issue

This question is about efficiency of dialog process compared to inefficiency of the background process and I want suggestions from SAP Basis/BI experts on this.
We have observed that our background processes on BI server are not running as efficiently as we would like them to run.
For example
DSO activation:
When I use Dialog mode for DSO Activation with 3 parallel processes, I can activate a certain DSO with in 15 seconds. But the same DSO and the same data package when activated using background mode with 3 parallel process takes about 15 minutes.(there are plenty of background processes available on the system when this activation is running). So dialog process runs 60 times faster than background process.
In BI 7 most of the operations can be executed only in the background parallel mode. And from the activation example we know that our background processes are not as efficient as dialog processes. It
is understood that in general the background processes will not be as efficient as dialog processes, but in our case the difference is a factor of 60. We want help to identify, why the background processes are not as efficient as dialog processes. What SAP Basis settings need to be changed to make them as efficient as we can. Any suggestion or help will be highly appreciated.

Hello Eswaran
Generally a dialog process is not faster by any means than a background workprocess. Nevertheless i trust you on your observations, so there must be some difference. I just did a small test, i ran SE16 chose a big table and selected 10000 rows, i did 3 tries in dialog and 3 in background. The results:
dialog: 10.2s, 9.4s, 9.6s
background: 26s, 24s, 24s
So even in my simple example the background execution took more time. The issue here is, that the resulting output (which is pretty large) had to be saved additionally in a spool (a total of 167 pages) when executed in background. The selection of the data certainly did not take more time in either case.
The most important difference between dialog and batch processes is the memory management. Dialog processes work with a shared memory segment (extended memory). Background processes have their private heap memory.
Nevertheless, background processes are not "slower" in general, not at all. You will need to observe closer, what your processes are doing. Maybe watching SM50 while running the DSO activation is enough to see it, maybe you need to trace a process. If you see large spools generated, or huge amounts of memory consumed, we might already have an answer.
Regards
Michael

Similar Messages

Line Split Interface to feed Integration Process - Performance Issues

Hi All
We have a scenario whereby we receive an XML message from a 3rd Party through an exposed SOAP Adapter service. The XML Message has multiple lines that need to be split up and processed as individual messages. We need to create a Line Splitting interface in order to achieve this. The Line Split interface would feed different Integration Processes depending on a specific payload value. The Integration Processes would then perform certain specific logic & Rules as well as transformations to specific message formats (e.g. idoc, xml, flatfile). The Line Split interface also maps from an xml structure that caters for multiple lines, to a flatened xml structure which only contains one line. The uper range of a message we may need to split into individual messages is 30 000 lines.
We first used an Interface Map and used SplitByValue to achieve this, however we ran into the constraint that we could not feed the output split messages to an Integration Process - you can only feed it to Adapters that reside on the J2EE engine.
We then decided to build a seprate Integration Process thats sole purpose was to split the message and route the indvidual messages to other integration processes to perform the logic, business rules and specific transformations. However, the performance of the ccBPM line splitting Integration Process was nowhere near the Interface Map.
e.g. Interface Map Split 1000 Lines = 13 seconds - BPM Integration Process 1000Lines = 100 seconds.
Does anybody have any suggestions on how we can perform the line split outside of BPM, or how we can improve ther performance of the line splitting within BPM?
Thanks for your assistance.
Edited by: CostaC on Aug 24, 2009 11:53 AM

hi,
>>>We first used an Interface Map and used SplitByValue to achieve this, however we ran into the constraint that we could not feed the output split messages to an Integration Process - you can only feed it to Adapters that reside on the J2EE engine.
the easiest (not the only) way :
do the split as you did here and post the results in different folders (file adapter)
then set up scenarios that will get the files from those folders
(many additional objects but will be much much faster and better then a BPM)
you could also split the messages in the adapter module but this is more advanced
and officially SAP does not recommend it - even though it's possible
Regards,
Michal Krawczyk

Synch Process Performance issues.

Hi,
I have a BPEL process that reads a file and then for each transaction within the file it calls a Synch BPEL process that processes it. The Synchronous process takes about 2 min. So say, if there are 25 transations in the file, it takes about 50 min to complete the entire processing. Earlier the Synch process was Asynch process but we had problems when all the records were processed at once that caused data duplication issues. Hence we made changes to make it a Synch process but then performace is become a problem.
Can we improve the performance in anyway?
Otherwise, if I make changes just have one process and move the entire logic of Synch process to the main one and process each record in a while loop, will that have better perfrmance than the above.
Need suggestions. Thanks!
-Prapoorna

We are on 10.1.3.4. So can't try on 10.1.3.5.
And we are using a specific schema.
Regarding the duplication, this is an EDI BPEL Process, so what was happening was, when we get the file and if the customer mentioned in the file does not exist in our system (Oracle db), then within the BPEL process we call the db adapter to call a custom procedure that creates the Customer. Because the process was Asynch earlier, if there are 4 transactions in the file, the Asynch process used to get invoked 4 times all at the same time and then we would see 2 or 3 customers created in the system with the same name and then a couple of instances would fail in the console when it is trying to find out if the customer already exists because the query returns multiple rows. Thats when I thought that may be I have to make the process wait until the 1st record is completely processed that will create the customer record if it does not exist and then when the 2nd record is processed it will not create the customer again as it is already created. Hence I made it a Synch process.
As I said earleir, do you think I can merge these two process (the one that reads the file and calls the Synch process for each record) into one process with a while loop... Will that work faster. appreicate if you cna give me suggestions on what my options are.
Thanks
-Prapoorna

Image Processing Performance Issue | JAI

I am processing TIFF images to generate several JPG files out of it after applying image processing on it.
Following are the transformations applied:
1. Read TIFF image from disk. The tiff is available in form of a PlanarImage object
2. Scaling
     /* Following is the code snippet */
     PlanarImage origImg;
     ParameterBlock pb = new ParameterBlock();
     pb.addSource(origImg);
     pb.add(scaleX);
     pb.add(scaleY);
     pb.add(0.0f);
     pb.add(0.0f);
     pb.add(Interpolation.getInstance(Interpolation.INTERP_BILINEAR));
     PlanarImage scaledImage = JAI.create("scale", pb);3. Convertion of planar image to buffered image. This operation is done because we need a buffered image.
     /* Following is the code snippet used */
     bufferedImage = planarImage.getAsBufferedImage();4. Cropping
     /* Following is the code snippet used */
     bufferedImage = bufferedImage.getSubimage(artcleX, artcleY, 302, 70);The performance bottle neck in the above algorithm is step 3 where we convert the planar image to buffered image before carrying out cropping.
The operation typically takes about 1120ms to complete and considering the data set I am dealing with this is a very expensive operation. Is there an
alternate to the above mentioned approach?
I presume if I can carry out the operation mentioned under step 4 above on a planr image object instead of buffered image, I will be able to save
considerable processing time as in this case step 3 won't be required. (and that seems like the bottle neck). I have also noticed that the processing
time of the operation mentioned in step 3 above is proportional to the size of the planar image object.
Any pointers around this would be appreciated.
Thanks,
Anurag
Edited by: anurag.kapur on Oct 4, 2007 10:17 PM
Edited by: anurag.kapur on Oct 4, 2007 10:17 PM

It depends on whether you want to display the data or not.
PlanarImage (the subclass of all renderedOps) has a method that returns a Graphics object you can use to draw on the image. This allows you to do this like write on an image.
PlanarImage also has a getAsBufferedImage that will return a copy of the data in a format that can be used to write to Graphics objects. This is used for simply drawing processed images to a display.
There is also a widget called ImageCanvas (and ScrollingImagePanel) shipped with JAI (although it is not a committed part of the API). These derive from awt.Canvas/Panel and know how to render RenderedImage instances. This may use less copying/memory then getting the data as a BufferedImage and drawing it via a Graphics Object. I can't say for sure though as I have never used them.
Another way may be to extend JComponent (or another class) and customize it to use calls to PlanarImage/RenderedOp instances directly. This can hep with large tiled images when you only want to display a small portion.
matfud

Virtual chars performance issues ?

Hello BW Experts,
Does virtual chars have performance issues ?
It has small effect or an adverse effect ?
please suggest.
Thanks,
BWer

hi BWer,
it depends on the 'logic' in virtual chars exit, may have performance issue e.g if you select/populate from
table, but may not if only filled with other characteristic value or only simple logic.
the 'how to' doc sample from sap market place have no performance issue.
hope this helps.

RKKBABS0 Performance Issues (Background Processing of CO99) for PM Orders

We are experiencing extremely long run times when batch processing through program RKKBABS0 in ECC 6.0 (just upgraded). The issue appears to be that the program is using the production order numbers to search against the EXKN table which contains no AUFNR or AUFPL information.
Has anyone experienced this same issue and how was it resolved?
Edited by: Ken Lundeen on Apr 9, 2010 9:17 PM
Edited by: Ken Lundeen on Apr 9, 2010 9:17 PM Table ESKN

(I'm sorry you've waited over a year for a reply.)
We also have performance issue. In our case we do not use Service Entry sheets with maintenance or production orders; AUFNR will not be populated in table ESKN. We are unable to 'complete business' our maintenance and production orders using batch processing because of performance.
We use Oracle database, which uses full table scan in this situation. But Secondary index (MANDT and AUFNR) is of no value anyway, we have about 12 million records with client and blank AUFNR field.
Our solution is a combination of a modification and a new index. OSS pilot note "1532483 - Performance of RKKBABS0 CHECK_ENTRYSHEET when reading ESKN" is a modification which introduces code improvements especially if running in background and closing several orders. Because we only have one client, we also created a new index consisting only of AUFNR. Oracle will not add a row to the secondary index of all fields of the index are null, making our new index very small. We then udpated Oracle stats to ensure Oracle would choose our new index.
We can now 'complete business' a single order online in under a minute, and the batch program runs much more efficiently.
This is not a perfect solution, but it has been a useful workaround for us. I hope this is useful to you.

Performance issues-assigning background jobs to a designated server

Hello all, we have been having peformance issues in our production enviorment and were entertaining the idea of running all our our background jobs on a separate/designated server. Has anyone had any experience with this and if so any insight on how you did this would be appreciated? Thanks.

I suppose your performance issues are then for the online users? In general, the best option should be to distribute the workload as much as possible, to avoid performance problems for online users as well as background jobs.
I agree with Atish, this is something for Basis people. Ask them (hopefully that's not you to look at their application server settings. I know where I used to work before they had set up the system so only a certain part of the dialog processes could be occupied by background jobs. I am sure SAP has a lot of parameters that can be tweaked, possibly also with time-dependent settings?

Heavy performance issues using Adobe Interactive Form PDFs generated by SAP BPM

Dear experts,
we use Adobe Interactive Form PDFs (generated with LiveCycle Designer) as Human Tasks within SAP BPM processes. The PDFs are generated and transmitted correctly, but when they are opened at the receivers PC, Windows freezes for 2-3 minutes, then the PDF opens and can be filled out and sent back. The next PDFs can be opened much faster, but when the PC is restarted, we get the same problem again. We use Adobe Reader XI (11.0.2) on our clients; is their any know performance issue?
Please note, that we have this problem with EVERY Adobe Interactive Form PDF... I created a simple PDF containing just a field and the client PC still freezes. So it can't be in the form or the scripting. Normal static PDFs can be opend without any problems.
Best regards,
David

They haven't really announced it, because there is no product to announce. Rather the opposite.
There are no conversion tools, so far as I know.
XFA forms are a non-starter if you want portability.
AcroForms are a nightmare in themselves, because the functionality is limited in Adobe Reader and varies between absent and weird in other products. No idea about Blackberry support.
You will not find a simple recommendation. Rather, you need to use Acroforms and carefully test everything (EVERYTHING: no assumptions) on every platform you intend to support.
Yes, rather unsatisfactory, but until Adobe realise that the future is platform equivalence or irrelevance, this is where we are.

Performance Issue with RF Scanners after SAP Enhancement Pack 5 Upgrade

We are on component version SAP ECC 6.0, and recently upgraded to Enhancement Pack 5. I believe we are on Net Weaver 7.10, and using RF scanners in one plant that is Warehouse Managed. Evidentially when we moved to EHP5, the Web SAP Console went away and we are left with ITS Mobile. This has created several issues and continues to be a performance barrier for the forklift drivers in the warehouse. We see there is a high use of java script, and the processors canu2019t handle this. When we login to tcode LM00, on a laptop or desktop computer, there are no performance issues. When we login to tcode LM00, with the RF scanners, the system is very slow. It might take 30 seconds to confirm one item on a WM Transfer Order.
1.) Can we revert back to Web SAP Console, after we have upgraded to EHP5?
2.) What is creating the performance issues with the RF Scanners now that we switched over to SAP ITS mobile?
Our RF Scanners are made by Intermec, but I donu2019t think that is where the solution lies. One person in our IT Operations has spent a good deal of time configuring SAPITS to get it to work, but still it isnu2019t performing.

Tom,
I am sorry I did not see this earlier.
I'm currently working on a very similar project with ITS mobile and the problem is to accurately determine the root cause of the problem in the least amount of time. The tool that works is found here: http://www.connectrf.com/index.php/mcm/managed-diagnostics/
Isolating the network from the application and the device is a time consuming process unless you have a piece of software that can trace the HTTP transaction between host and device on both wired and wireless side of the network. Once that is achieved (as with Connect's tool) you can then you can begin to solve the problem.
What I found in my project is that the amount of data traffic generated by ITS mobile can be reduced drastically, which speeds the response time of the mobile devices, especially with large number of devices in distribution centers.
Let me know if I can answer more questions related to this topic.
Cheers,
Shari

Background processes on SAP BW Production stall - kicked off via UC4

All,
We have the following issue; there 6 background processes available on the BW server. The UC4 (external) scheduler kicked off 3 jobs that in turn was kicking off 3 process chains which were trying to kick off 10 background processes. The result was that 6 background processes were kicked off, seemed to be active (in sm37), but actually stalled (they were not doing anything), it was like they ended up in a deadlock. Normally one would expect the processes to run according to the mechanism that when there are no background processes available, to wait. My guess is that because they were initiated by an external scheduler and demanded more than the total available background processes that mechanism did not work. Has anyone encountered this problem before? And knows why it happens?
Regards, Meindert

Hi Meindert,
If I recall correctly I remember having seen this before as well, and then the root cause actually was the way the process chains work. There was (is?) some dependency on free batch processes in the initialization phase of the process chain which lead to hanging processes if you started a few process chains at the same time and there are not enough free batch processes available.
I believe this is internal to SAP so probably not caused by scheduling the chains externally, however your external scheduler could maybe take the number of free batch processes into account. I do not know your external scheduler well enough to explain it for that product, I do know how I would handle it with CPS: I'd use queueing to make sure that there are enough free batch processes available, or I would let it monitor the number of free batch processes on the system and if that would become too low, I'd temporarily hold new process chains that needed to start. The chosen option would depend on the priorities of the processes involved, to make sure that the processes with the highest priorities can always start and always have batch processes available.
Regards,
Anton.

Performance issues with SAP BPC 7.0/7.5 (SP06, 07, 08) NW

Hi Experts
There are some performance issues with SAP BPC 7.5/7.0 NW, users are saying they are not getting data or there are some issues while getting data from R/3 system or ECC 6.0. Then what things do I need to consider to check, such as what DataSources or Cubes I need to check? So, how to solve this issue?
What things I need to consider for SAP NW BI 7.0 u2013 SAP BPC 7.5 NW (SP06, 07, 08) Implementation?
Your help is greatly appreciated.
Regards,
Qadeer

Hi,
New SP was released in February, and now most of the new bugs should been caught ,This has a Central Note. For SP06 it's Note 1527325 - Planning and Consolidation 7.5 SP06 NetWeaver Central Note to fix any issues. Most of the improvements in SP06 were related to performance, especially when logging on from the BPC clients.There you should be able to find a big list of fixes/improvements and Notes that describe those. Some of the Notes even have test description how to reproduce that issue in the old version.
hope this will help you
Regards
Rv

Performance Issue in Dashboard using SAP BW NetWeaver Connection

HI Experts ,
We developed Dashboard which is based on BW Queries, However it is taking considerable amount time while executing.
We are using BO Dashboard SP2 version ,BO 4.1 and BI system 7.01 .
We are looking for few clarifications on SAP BO Xcelsius Dashboards.
Though we know limitations on component number and data volumes which could badly affect performance of the dashboard, We do have a requirement to handle huge data volumes and multiple components. Our source data lies in SAP BI system and we are using BICS connectivity/ Webi with QAAWS for updating data in BO dashboard .
Our requirement is too complex where we should be in a position to meet user expectations for complete view 75 KPI’s in a single Dashboard.
We have scenarios like YTD ,QTD and MTD other complex calculations in Bex Query.
Here are my questions,
Is there any way to provide complete functionality using large data sets to the users with the current architecture without any performance issues?
Are there any third party tools which can be used with BO Dashboard for the performance improvement and handling huge volumes?
Do you suggest any alternate solution for complete functionality?
Many thanks for your inputs in advance!
Regards
Venkat

Hi Rajesh,
Thank you so much for your response.
I have tried the way you suggested. But here my issue is , I have a prompt in my webi report based on months selection and it is a single valued prompt.
so I was able to schedule my report only for one month;whereas my dashboard needs to show values for one year data based on the months the user will select in the dashboard.
Though i use the WebI instance data in the dashboard , i getting only one month value, also I am not able to associate the selected month for the WebI instance in dashboard.
Is there any option to schedule webI report for different months and the dashboard has to pick the 12 months instance and the combo box selection in the dashboard must associate with it??
Please help me with your thoughts.

Performance issue in Webi rep when using custom object from SAP BW univ

Hi All,
I had to design a report that runs for the previous day and hence we had created a custom object which ranks the dates and then a pre-defined filter which picks the date with highest rank.
the definition for the rank variable(in universe) is as follows:
<expression>Rank([0CALDAY].Currentmember, Order([0CALDAY].Currentmember.Level.Members ,Rank([0CALDAY].Currentmember,[0CALDAY].Currentmember.Level.Members), BDESC))</expression>
Now to the issue I am currently facing,
The report works fine when we ran it on a test environment ie :with small amount of data.
Our production environment has millions of rows of data and when I run the report with filter it just hangs.I think this is because it tries to rank all the dates(to find the max date) and thus resulting in a huge performance issue.
Can someone suggest how this performance issue can be overcome?
I work on BO XI3.1 with SAP BW.
Thanks and Regards,
Smitha.

Hi,
Using a variable on the BW side is not feasible since we want to use the same BW query for a few other reports as well.
Could you please explain what you mean by 'use LAG function'.How can it be used in this scenario?
Thanks and Regards,
Smitha Mohan.

Performance issue of BI reports in SAP Enterprise portal

Dear Friends,
We have integrated BI reports with SAP Enterprise portal 7.0.Reports are running properly But the issue is reports are taking more time to dispsaly its content and leading it to performance effect.
In Bex ( BI side) reports performance is little better than SAP EP platform. BI Team also looking for ways to improve performance at BI side.
Could you please share your valuable ideas to improve the performance at SAP EP side also ..
Thanks and Regards
Ratnakar Reddy

Hi ratnakar,
The first step is to identify which component is causing the performance problem. Run your report in the portal but try appending the string &PROFILING=X in the end of the URL. This will generate BI statistics which you can use to see which component (Java stack, ABAP stack, Database) is causing the performance issue.
Hope this helps.

Performance issue and data getting interchanged in BO Webi report + SAP BW

Hi,
We are using SAP BW queries as the source for creating some BO reports.
Environments :
SAP - SAP BI 7.1
BO - BO XI 3.1
Issues :
The reports were working fine in Dev and Q with less data. But when we point the universes to BW prod ( where we have much data), the reports are taking quite a long time to refresh and getting timed out. This query has some key figures which are having customer exits defined to show only one month data. And also BW accelerators are updated for the infocubes pertaining to this query. The BO report is giving data if we apply a filter in 'Query Panel' of Webi to show only current month dates. But then the issue is the values are getting interchanged for many objects. For ex: there are 2 objects- ABS version and Market region. The values are getting interchanged in the BO level.
Please let us know if anything needs to be done in BO or BW to fix this issue if anyone has faced the same
Also Please let us know if customer exits and accelerators works fine with BO
Thanks
Sivakami

Hi,
Thanks Roberto. We'll check the notes
@Ingo,
We are able to solve the performance issue by removing unused Key figures and dimensions from the query, but the column value interchange issue still persisits
The build version is - 12.3.0
Query Stripping
Where should we enable query stripping? When i went through some documentation it was written that it'll be enabled automatically from XI 3.1 Sp3. Can you please let us know if its so and what we need to do to enable it.
The coulmn interchange is happening when we use dimensions in a certain order. When product type is used along with Market region. Market region shows values of Product type also in Webi report.
Thanks & Regards,
Sivakami

SAP BI 7 - Background process performance Issue

Similar Messages

Maybe you are looking for