Code highlighting

Showing posts with label Performance. Show all posts
Showing posts with label Performance. Show all posts

Thursday, March 02, 2017

Announcement: Warehouse Mobile Devices Portal improvements with February update of AX 2012 R3

A monthly cadence AX 2012 R3 update for February has just come out on LCS, and with it a few changes our team has done that deal with overall behavior and performance of WMDP - the Warehouse Mobile Devices Portal used in the Advanced warehousing module.

I want to call them out here, and if you are reading my blog to keep up to date with the Warehousing changes, I strongly encourage you to install the below changes:


  • A functional enhancement, that allows you to start execution of a work order that has some lines awaiting demand replenishment, processing the lines that can be picked already now
  • A performance optimization, that avoids updating certain persisted counters on the wave when work goes through its stages. 
    • This also ensures they do not get out of sync, as they were used to make certain decisions about what is allowed for a work order
    • Can also be downloaded individually as KB3217157
  • An integrity enhancement, that ensures we always are in a valid state in DB, where each service call from WMDP is now executed within a single transaction scope
    • Can also be downloaded individually as KB3210293
    • Can be turned off in code for a selected WHSWorkExecuteDisplay* class, if needed
    • This is a great change ensuring we do not commit any data unless all went well, but might hypothetically impact your new/modified WMDP flows, if you handle your exceptions incorrectly today. If you do find issues with them, please let me know, I am curious to know your specific examples.
  • Various minor enhancements, that ensure WMDP performs well under load
    • Can also be downloaded individually as KB3210293
    • This includes improvements to enable better concurrency in various WMDP scenarios, better error handling on user entry, etc.

Again, install these, try them out, and provide feedback!

There are a lot more enhancements and bug fixes that went into this release, you can read the full list by following the link below:


Thanks!

Monday, May 11, 2015

Tutorial: Caching display methods

Overview

Caching is an important part of the Microsoft Dynamics AX application. It allows us to significantly speed up certain processes by avoiding redundant round trips to the server and the database. Display/Edit methods are a good example of where caching is helpful, as it significantly increases the overall responsiveness of a form.
In AX 2012 R3 there are a number of ways you can specify that a display method needs to be cached. Below is a list, as well as a brief description and an xpo with a tutorial form to show the various options.
  • Invoke cacheAddMethod(MethodName) in the init() method of the data source
This is the only option that was available in older versions of Dynamics AX.
All cached data methods are calculated and cached in executeQuery of the data source they belong to. That is why it is important to call the above method in the init() of the data source, once the data source is initialized, but the data hasn't been retrieved yet. The optional second parameter controls whether or not the display method value should be refreshed when data is written to DB
  • Add the SysClientCacheDataMethodAttribute attribute on the display method
Now that attributes are available, caching a display method is as easy as adding an attribute on the method. It has an optional parameter same as the above method. Note, that marking a display method with this attribute will mean that it is cached on all forms the method is used on. So make sure that is what you want to happen before adding it.
  • Set the “Cache data method” property on the display method form control to Yes
This is, again, a new capability in the latest version of AX, that is pretty much identical to calling the cacheAddMethod, just much easier.
  • Set the “Cache data method” property on the display method form control to Auto, where the data source is read-only
Now, this is something that is done automatically by the kernel. If the data source which the display method belongs too is read-only, and Cache data method property on the corresponding control is set to Auto, the kernel will cache this method. This is smart, as the value can’t really change. You can still change the property to No, if there’s a reason why you want the display method not to be cached.

Additional info

On top of the above, I would also like to mention a few other things.

  • There are 2 kernel-defined methods, called titleFields and parentTitleFields, which are always being cached automatically by the kernel. The titleFields method corresponds to the 2 Title Fields that you can set on the table level. Then, if you have this table as your TitleDatasource on the form design, the 2 fields selected will be shown in the caption of the form. parentTitleFields is very similar, only it displays the title fields of the parent data source table, which is relevant in a header-lines (dynalinked datasources) kind of scenario. As a fallback, it shows the titleFields.
  • There are methods for managing the cached methods, one of them being cacheCalculateMethod, which asks the server to re-calculate the method value and update the cache for the current record. Note that it will only take effect if the method has previously been cached. In the tutorial project I am using this method in the form method runHack to refresh the value of 2 of the display methods. 

Example

You can download the xpo project with the tutorial from my OneDrive.
Upon opening the form in the project, some data will be pulled into the form. More specifically, the first 3 items/customers/vendors/sites/warehouses. For each of them a display method is present, and one of the above approaches is used to cache it.

Step 1



None of the display methods are cached. As you can see, nothing extra was done in executeQuery of the main data source, but itemName was cached for all 3 records for the read-only data source, as the control has "Cache data method" property set to Auto. This happens even though the tab page is not the active one, so it wasn't really yet necessary to calculate these values. Maybe the user won't even navigate to that tab page. You should consider this carefully when designing a form, and always aim at increasing the overall responsiveness of a form by only executing the appropriate display methods.

Step 2 and 3



inventSiteName and vendName were cached (through one of the options above). As you can see, now their values are calculated once for each of the records shown. InventLocationName and custName are still not cached.
Clicking on Refresh result button will show any extra calls to the display methods since the form started. As you’ll quickly notice, the non-cached methods are being called every time something hints at covering up the value of the control, or rather every time AX thinks it needs to recalculate the value, which is VERY OFTEN. That should make it pretty clear how useful caching is for display methods.

Step 4



Now inventLocationName was also cached, and the only method that is remaining and not cached is custName. You can see how more and more methods are executed in one batch on the server as part of the executeQuery method on the data source.

Step 5



As the last step, I have added 2 standalone controls that show the Title fields of the main data source table, and started logging the time method caption is invoked. As I have written above, titleFields and parentTitleFields are cached automatically, but you can't really see it, as it is not possible to override it. caption is the closest you can get to it, so I figured it's worth showing that. You can see the method is evaluated right after all data source queries have executed (as it relies on the currently selected record). It is evaluated first time based on the selected data source record, and then again when focus actually goes to the grid, so that the first record is explicitly selected. It will be invoked again whenever the active record changes. Note that unlike custName method, it does not get constantly refreshed.

Play around with the form and let me know if you have any questions or comments.
Thanks

Friday, October 28, 2011

Tutorial: AX 2012 - Invalid field access or Accessing unretrieved fields

Read the post all the way until the end - there's an ask for you guys there!

As many of you have already seen in the new AX 2012 release, there are a lot more tables present in the application, because we went through an exercise of normalizing the data model.
Another feature that was added at the same time was the Table Inheritance, allowing to sub-class tables and add additional fields reusing some of the behaviors of the base table.

All of it is great! The only problem with it is that it comes at a price - we now need to retrieve a lot more data and do a lot more joins between tables. So we tried to mitigate that by using field lists where possible and adding AdHoc query support, which basically eliminates unneccessary joins between tables from the hierarchy, if fields from these tables are not selected.

We have also implemented a number of things that allow us to clearly see if the field was selected in the user interface, the APIs needed for doing the same, as well as special handling for invalid field access.

This tutorial contains a form, which showcases the different aspects of data access in cases described above.
You can download the project with the tutorial from my SkyDrive.

On the first tab, we have the standard "On hand" view, which in the data model is a join between InventSum and InventDim, with group by clauses on selected fields, in this case, ItemId from InventSum and InventLocationId from InventDim, and aggregation on the AvailPhysical column.
As you can see from the screenshot below, the rest of the InventDim fields are shown as Unretrieved. So, naturally, accessing one of them from code, for example, should not be a legal operation.
To verify that statement, there are 2 buttons on the form, the Incorrect and the Correct was of accessing the field. Clicking on the first one simply tries to read the value out of the InventSiteId field, while the second one uses the API method TableBuffer.isFieldDataRetrieved() to first verify if the field can actually be accessed.

Note, that in the below example, both buttons will work, as in, there won't be any errors shown to the user. When accessing the field, even though it is not retrieved to the client, the value will be treated as the default value for that type, that is, an empty string.

In order to verify that we do not access fields in an invalid way like above, we have introduced a parameter, that will throw an exception if a field that was not retrieved is being accessed.

In order to enable this validation, you need to update the below shown parameter in the Server configuration form (Located under System administration \ Setup \ System). After changing the value (note, that it is per AOS) you need to restart the AOS for the changes to take effect.


Now, if you try to use the first button (under Incorrect) from above, you will get a stack trace, notifying you that the specified field was not retrieved.



We suggest that when testing your modifications in the application, you always have that flag enabled, so as to avoid unpleasant and hard-to-find bugs later on in the production environment.


On the second tab of the Tutorial form, the same type of information is presented, only this time the output is actually based on a new table inheritance structure I created.


The base table, GenericBall, contains 2 characteristics of any ball. SoccerBall is extending it and adding an additional characteristic (for the sake of the example, let's assume Brand is only relevant to soccer balls). This is basically a very simple table inheritance structure.

In the form, however, I am only selecting to view the Brand field from SoccerBall, not selecting the other 2 fields from the base table. As expected, they show up as Unretrieved in the user interface.
However, there is one difference in how Scsc tables are handled - and that is, they will always throw an exception when you try to access one of the unretrieved fields.
The two buttons above the grid demonstrate that. Try it out, enabling/disabling the server configuration parameter shown above.


That's pretty much all there's to it. Let me know if you have any questions.

Now, I have one thing to ask you all too.
We in the AX Test team have done our best to find invalid field access problems before the release, but if you find one using the approach above, please log it either through the standard Microsoft channels (partner/ MsConnect/ etc.), or as a comment directly under this blog post.

Thank you!

Sunday, February 07, 2010

Performance optimization: Deleting inventory journal lines

Preamble

As a developer, you should always consider performance implications of the code you write. In an ERP application like Microsoft Dynamics AX, the main focus should be on query execution, since it takes up the overwhelming part of the servers' resources.

You should always write queries that would execute the minimum amount of time and use the minimum amount of resources, at the same time producing the expected output in all cases.

Note, that performance is one of those things you cannot really verify on a 1-box install with a small test dataset. Most query problems show up only when tested with many users concurrently loading the AOS on a large-size database.

Code example

Table method
AOT\Data Dictionary\Tables\InventJournalTrans\Methods\delete
contains the following code:

if (this.Voucher)
{
if (this.numOfVoucherLines() == 0)
JournalError::deleteVoucher(tablenum(InventJournalTable),this.JournalId,this.Voucher);
}

The code is logically correct, deleting related records from the JournalError table, which contains error messages generated during validation and posting of journals.
But now let us consider the actual implementation. If we rephrase the conditions under which we delete the error message history, it would sound something like:
If a Voucher number is specified on the line being deleted, and there is no more lines in this journal that use this Voucher number, then we remove the JournalError records.

Deeper code analysis for performance

So, let's go in deeper, and open the code for method numOfVoucherLines:

Integer numOfVoucherLines()
{
return any2int((select count(RecId) from inventJournalTrans
where inventJournalTrans.JournalId == this.JournalId &&
inventJournalTrans.Voucher == this.Voucher).RecId);
}

As you can see, it counts all the journal lines with the specified Voucher number and JournalId.
Does it generate the expected output? Yes.
Is it optimal for the scenario in which it is used? No.

In order to determine if JournalError records can be deleted, we only need to know if at least 1 record with the specified Voucher number exists. So why do we need to search and count all such records? We don't.
Note that this code is being executed for each journal line, magnifying the impact of a non-optimal query N-fold.

Suggested solution

So, if we create a new method like this:

boolean voucherLineExist()
{
return (select firstfast firstonly forceplaceholders recId from inventJournalTrans
index hint VoucherIdx
where inventJournalTrans.journalId == this.journalId &&
inventJournalTrans.voucher == this.voucher).recId != 0;
}

and use it instead of numOfVoucherLines() in delete() method, we will greatly improve the overall performance when deleting journal lines.
It has been measured, that the time it takes to delete a journal with a large number of lines (88000, to be specific) has reduced by 6 times, which is fantastic.

Notes

  • forceplaceholders is used in the query explicitly to make sure a query plan is created and reused for this query, which further improves the performance, since the Voucher number most probably changes from line to line (depends on related journal name setup).
  • It is generally still a best practice to avoid creating inventory journals with such a large number of rows, because the journal is usually processed in 1 database transaction.

See also