Schedulability Analysis - Any thoughts?

oahmad2 wrote on Thursday, April 15, 2010:


I was reading up a little bit of theory on Real Time Systems and I saw mention of schedulability-analysis that is done ensure that tasks meet their deadlines.  I saw the equation for RMS and was wondering, in a practical sense, do the forum members do this for their projects, and how do they do this.  Does anyone have any practical experience or examples that they can share on how to do this best in practice?  For example, does FreeRTOS have any mechanism to profile the different tasks’ execution time or is this usually done manually using a toggled bit and oscilloscope?

Lastly, Off Topic: but is there an easy way to search the forum? 


edwards3 wrote on Thursday, April 15, 2010:

I think you will find that most people do not bother with a formal calculation unless there is a safety critical aspect to their application, or they have a large volume mass market product and they want to use the least powerful cpu possible.

Schedulability is an easy exercise in an academic environment, and extremely difficult in a real environment. Often it relies on all tasks being periodic, which they rarely are. Also once you start adding in interrupts or any other asynchronous events then all the calculations fall apart.

There are other schemes that allow easier analysis. For example, time triggered systems, or scheduling algorithms that partition time between tasks. These are less flexible and wasteful of cpu cycles but offer better predictability.

edwards3 wrote on Thursday, April 15, 2010:

To answer the other part of your post, have you seen these pages and ? Very powerful trace schemes can be created with very little effort.