Java Threads, Second Edition

Round-Robin Scheduling

Our next examples show two ways of performing round-robin scheduling. These techniques are mostly useful for programs that will execute on a green-thread implementation of the Java virtual machine, since native-thread implementations all perform some sort of round-robin scheduling already. If you don’t know which implementation your program will eventually run on, you may want to use one of these techniques to ensure round-robin scheduling, even if the underlying implementation is performing it for you: these techniques are not in conflict with native-thread implementations, though they do not necessarily provide a benefit on those platforms.

Remember that at issue here is the behavior of a Java program that contains one or more CPU-intensive threads. A Java program could have hundreds of threads that may only periodically need access to the CPU and otherwise spend most of their life in the blocked state: in that case, there isn’t likely to be much competition for the CPU, and each thread gets an opportunity to run whenever it has work to do. We only face the problem of CPU starvation when there is at least one CPU-intensive thread that may potentially prevent all other threads from running.

If we have only a single CPU-intensive thread, there is no need for a complicated scheduling mechanism: all we need to do is lower the priority of the CPU-intensive thread below the priority of the other threads in our Java program. This allows the other threads to ...