Hi.  This is my first ever web log post, so please be understanding while reading it. :-)

I recently converted a block of test scheduling code from C++ to C# and decided to choose a new design paradyme.  I was quite pleased with the results. 

The C++ code used a list of tests that was unordered and a dynamic pool of threads that consumed the tests to execute them.  There was a large amount of code to handle the thread synchronization, the dynamically growing thread pool and support for generic test execution.

The new C# code used two collections of tests and takes advantage of the built in thread pool in the .Net Framework.  The basic functionality is:

1. Tests get added to the "pending" collection, which sorts them as they are added, by the time they are due to execute. 
2. A timer is set to go off when the next test is due.
3. When the timer expires the tests that are due are moved to an "executing" collection, which spins up a worker thread for each test, and a timer to track the test in case it times out.
4. When a test completes, its results are stored in a "result" collection and the test is moved back to the "pending" collection.

The whole process it self perpetuating and this design really simplifies the code and makes it much more maintainable.  I was also amazed at how much smaller the resulting C# code was compared to the C++ original.  I'm sure the .Net Framework is largely responsible, but from a developer's standpoint, I'm still very pleased. 

Howitzer