PxPlus User Forum

Main Board => Discussions => Language => Topic started by: michaelgreer on October 02, 2019, 04:20:27 PM

Title: File performance
Post by: michaelgreer on October 02, 2019, 04:20:27 PM

I have a customer whose history files have years (10+) of history.  Will this degrade overall performance. I process that they run daily is now taking a good deal longer than it used to and we are looking at ways to resolve this.

Title: Re: File performance
Post by: Mike King on October 02, 2019, 05:12:54 PM

Obviously larger files take longer to read thru -- so depending on your logic larger files could increase the run time for some of your processing.

There are however a number of ways to improve performance.

• One thing you might want to look at is, if the files are standard VLR, changing the file block size.  Typically file have a 4K block which means the file is split into a number of 4K blocks each of which can contain keys or data. A small block can result in a limitation on the number of key entries per block, especially if you have small keys.  You might type increasing the block size to improve performance. 
• If running a current PxPlus try setting the 'XK' system parameter and reloading your files using a larger block size; this parameter allows the system to pack more keys into each block.
  For the process that they are having issue with, if it is typically run stand-alone, try LOCKING some of the files.  This improves file performance as the system does not have to worry about updated from other processes.
• Try increasing the number of buffers for the files using the NBF= option on the OPEN or SET_NBF to increase the global buffer pool.
• In the long running process, if you have files that you read only and don't update, try doing an OPEN LOAD on the files as this attempts to cache all or portions of the file in memory.

Lastly, make sure you processes attempts to use keys to filter the data.  I have seen instances where nightly processing simply reads through ALL records and then filters by date as opposed to reading using an alternate key that is based on the date required.

Now if all else fails, try to make sure the system is using SSD disc drives as these tend to run faster.