I recently had a customer ask me why Windows was showing a much larger memory footprint than what was expected (based on their cube sizes). After further analyzing their model, we saw a large amount of garbage memory … which led to the question “what is garbage memory?”
As you know, IBM Planning Analytics uses various approaches to minimize the amount of RAM needed for a model. PA (Planning Analytics) is very good about only taking what it needs to support views, calculations, and processing. However, I sometimes like to compare PA to a small child – it is very good at only taking what it needs, but it’s not so good at giving it back when it’s done. This is does this on purpose since results are stored for future reference, thereby providing faster results for your users. The excess memory that was being and is no longer needed is called garbage memory.
You can determine how much memory is being used by enabling performance monitoring and reviewing the }StatsForServer control cube.
So how do you release (or reset) the amount of memory used by PA? You can do this by simply recycling your PA server. Revelwood recommends recycling your service at least once per week. Not only will this reset the amount of memory used on the machine, but it will also help with other basic maintenance tasks such as ensuring your transaction log data is properly written to disk.
Revelwood has worked with IBM Planning Analytics / TM1 for more than 27 years. We’ve partnered with hundreds of companies on the design, development, maintenance and updates of IBM Planning Analytics applications, across every industry. Have a challenge with Planning Analytics / TM1? We can help you!