Versions Compared

Old Version 1

changes.mady.by.user Jenni Wallin

Saved on

compared with

New Version 2

changes.mady.by.user Former user

Saved on

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

...

The Shared Table profile is loaded when you start a workflow that depends on it. Changes to the profile become effective when you restart the workflow and each time you save the profile.

Memory Allocation

There are three different ways to allocate memory for the created tables. By default, the tables are kept as Java objects in memory. The shared tables can also be configured to keep the tables as raw data either on or off the heap. By using raw data, the overhead of java objects is removed and less memory is required.

...

For further information about memory allocation in, see the System Administration Guide.

Shared Table Profile Configuration

The Shared Table profile configuration is opened by clicking on the New Configuration button in Desktop and selecting the Shared Table Profile option.

...

SettingDescription

Database

Click the Browse...button and select the Database profile you want to use. Any type of database that has been configured in a database profile can be used. See 8.7 6 Database Profile for further information.

Release Timeout (seconds)

If this check box is selected, the table will be released when the entered number of seconds have passed since the workflows accessing the table were stopped. The entered number of seconds must be larger than 0.

If this check box is not selected, the table will stay available until the execution context is restarted.

Refresh Interval (seconds)

Select this check box in order to refresh the data in the table with the interval entered. The entered number of seconds must be larger than 0.

If this check box is not selected, the table will only be refreshed if the APL function tableRefreshShared is used. For more information regarding the function, see the section below, tableRefreshShared.

Note
titleNote!

The interval for checking if the table needs to be refreshed is 10 seconds, which is the minimum time before a new refresh is performed.

If a refresh fails, an error is generated in system log, but the table is not cleared - the old data remains in the shared table. A new refresh is initiated every 10th second, until the refresh has finished successfully.


Object

Select this option to set the Table Storage to Object. If you select this option, the shared tables are stored as Java objects on the JVM heap.

Note
titleNote!

If you have selected to use a profile with the CSV database type, this is the only option available if you have not configured properties using Advanced Connection Setup.


On Heap

Select this option to set the Table Storage to On Heap. If you select this option, the shared tables are stored in a compact format on the JVM heap. If you select On Heap, you must select an option for the Index Storage.

Note
titleNote!

If you have selected to use a profile with the CSV database type, this option is only available if you have configured properties for this using Advanced Connection Setup.


Off Heap

Select this option to set the Table Storage to Off Heap. If you select this option, the shared tables are stored in a compact format outside the JVM heap.

Note
titleNote!

You are required to set the jdk parameter in the relevant Execution Context pico configuration, for example:

Code Block
languagetext
themeEclipse
$ mzsh topo set topo://container:<container>/pico:<pico>/obj:config.jvmargs \
'maxDirect:["-XX:MaxDirectMemorySize=4096M"]'



Note
titleNote!

If you have selected to use a profile with the CSV database type, this option is only available if you have configured properties for this using Advanced Connection Setup.

If you select Off Heap, you must select an option for the Index Storage.

Unsafe

Select this option to set the Table Storage to Unsafe. If you select this option, the shared tables are stored in a compact format. If you select Unsafe, you must select an option for the Index Storage.

Note
titleNote!

If you have selected to use a profile with the CSV database type, this option is only available if you have configured properties for this using Advanced Connection Setup.


Primitive Lookup

Select this option to set the Table Storage to Primitive Lookup. This provides simple lookup tables with a fast lookup function but they are limited to two columns of type Int/Long for the key (column 1) and type Short/Int/Long for the value (column 2). Lookup operations on Primitive Lookup tables are limited to the equals operation on column 1.

Note
titleNote!

If you use the  Primitive Lookup  option with a database profile that is configured for Oracle, using the Oracle column type NUMBER with a precision greater than 10 may cause errors.


Note
titleNote!

If you have selected to use a profile with the CSV database type, this option is only available if you have configured properties for this using Advanced Connection Setup.


Object

Select this option to set the Index Storage to Object. If you select this option, the index is stored as Java objects on the JVM heap. This option is only available if you have selected On Heap, Off Heap or Unsafe for Table Storage.

Pointer

Select this option to set the Index Storage to Pointer. If you select this option, the index is stored as pointers to the table data. This option is only available if you have selected On Heap, Off Heap or Unsafe for Table Storage.

Cached Long/Int PointerSelect this option to set the Index Storage to Cached Long/Int Pointer. This option is only available if you have selected On Heap, Off Heap or Unsafe for Table Storage. For numeric index columns, the Cached Long/Int Pointer can be used for faster lookups, but at the cost of slightly higher memory consumption.
Variable Width Varchar ColumnsSelect this check box to enable variable width storage of varchar columns. This reduces memory usage for columns that are wide and of varying width.

SQL Load Statement

In this field, an SQL SELECT statement should be entered in order to create the contents of the table returned by the tableCreateShared APL function.

The following statement will return a table named MyTable with the columns key and value when the tableCreateShare function is used together with this profile.

Info
titleExample - SQL SELECT statement


Code Block
languagetext
themeEclipse
SELECT key,value FROM MyTable


If no data has been fetched from the database, SQL errors in the table lookup will cause runtime errors (workflow aborts). However, if data has already been fetched from the database then this data will be used. This will also be logged in the System Log.

Whenever possible, use values of the type long instead of the types date or string. This may improve performance.

Info
titleExample - Using SQL SELECT statement


Code Block
languagetext
themeEclipse
SELECT to_number(user_id) as USER_ID, 
 to_number(to_char(nvl(start_date,to_date(
	'19000101010101',
	'yyyymmddhh24miss')),
	'yyyymmddhh24miss')) as START_DATE,
 to_number(to_char(nvl(end_date, to_date(
	'99990101010101',
	'yyyymmddhh24miss')), 
	'yyyymmddhh24miss')) as END_DATE
 FROM MyTable



Table Indices

If you want to create an index for one or several columns of the shared table, these columns can be added in this field by clicking the Add...button and adding the columns for which you want to create an index. The index will start with 0 for the first column.

Note
titleNote!

An index will not be created unless there are at least five rows in the table.

Create indices in the Shared Table profile based on the data that is fetched from the database. Even if you look up multiple columns in APL, using one index instead of several may result in improved performance. 

...

Note
titleNote!

Some Database Management systems provide character column types so that you are not required to specify the column width (e g TEXT in PostgreSQL). If you use shared tables with such a column type, you cannot use the types On Heap, Off Heap or Unsafe within Table Storage.

APL

The following functions are included for the Table Lookup Service:

  • tableCreateShared

  • tableRefreshShared

tableCreateShared

This function returns a shared table that holds the result of the database query entered in the Shared Table profile.

...

Info
titleExample - Using the function tableCreateShared

This creates a shared table called myTable with the columns returned by the SQL query in the mySharedProfile Shared Table profile.

Code Block
languagetext
themeEclipse
initialize {
  table myTable = tableCreateShared("Folder.mySharedProfile");
  }


tableRefreshShared

This function can be used for refreshing the data for a shared table configured with a Shared Table profile. The table will be updated for all workflow instances that are using the table and are running on the same EC.

...

Info
titleExample - Using the function tableRefreshShared

This returns the shared table called myTable, which uses the mySharedProfile, with refreshed data.

Code Block
languagetext
themeEclipse
table myTable = tableRefreshShared("Folder.mySharedProfile");


Additional Performance Tuning

The Oracle JDBC driver includes a feature built that allows you to set the number of rows that are prefetched while the full result set is being formulated. At the time of writing, the default number of prefetched rows is 10. You can increase this value by setting the Execution Context property oracle.jdbc.defaultRowPrefetch in the relevant pico configuration file.

...