Performance Related Options

About Performance Related Options

These are the factors that improve the performance of DBLoad:

  1. The default EJS installation has the access log turned on in admin/configs/xjs.properties. This should be set to false.
    To improve performance turn off all the logging events.

  2. Set these entries in admin/configs/dbinfo.cfg.

  • enableperuserformcache=true

    Caches the view information per user if this field is set to true.

  • numberofcachedforms=100
     
    Enables the configuration of the number of XDForm cached in the server. There is a new flag in the dbinfo.cfg file called numbercachedforms with a default value of 100. This allows the server to be configured to have a different limit on the number of cached forms. Values of 1 or any positive integer are treated as their indicated values. Negative numbers or nonnumeric strings are converted to an internal flag(-1) that means an infinite number of forms can be cached. This value can be set to a higher number as appropriate to improve performance.

  1. Batch size

    Batch size refers to the number of instructions (items) to be processed in one batch for a database operation. The batch size can be reset by using the -b option on the command line. The default batch size of 1 (one) is used to cater to data files that have dependencies between items in the same file.

    These are the example of dependencies between items:

  • Cascaded edit requires the dependent object be edited subsequent to the items being depended on (for example, pointer relationship)

  • Self-referencing objects.

  • Items with notes and attachments processed in the same file, which requires the item to be inserted first before trying to insert the note and attachment.

    The batch size can be set to a higher value based on the size of the item and memory available on the system.

    Note: Higher performance can be achieved by increasing the batch size. But it is user's responsibility to make sure that the items contained in the files are not dependent and can be processed irrespective of their sequence in the file.

  1. Worker Size

    Worker size refers to the number of parallel transactions that are initiated by dbload. By default Dbload uses a worker size of one. The default batch size of one is used to cater to data files that have dependencies between items in the same file. When the worker size is one, a single thread processes the file in a sequential manner. You can reset the worker size by either making an entry in the dbload.cfg file or by using the -w option in the command line. The conditions for using higher worker size are the same as batch size.

Note: Higher performance can be achieved by increasing the worker size. But it is user's responsibility to make sure that the items contained in the files are not dependent on each other and can be processed irrespective of their sequence in the file.