Context-Based Synchronization of Concurrent Process Using Aspect-Oriented Programming

Resource access synchronization within concurrent processes or threads is implemented using various constructs such as semaphores, monitor locks, Mutex, etc. The algorithm supporting most of these structures works by keeping at bay all other concurrent processes or threads till the current process accessing the resource has successfully relinquished the resource. This works very well as race conditions are controlled and shared data state remains consistent. The problem with this approach is performance in terms of system response. When each thread has to wait for the other to finish accessing the resource before it can proceed, a long line waiting threads can easily build-up, which obviously translates to relatively slow system response. In this paper, we propose contextual synchronization model to avoid where applicable, the unnecessary build-up of threads waiting for access to the resource. This model describes different contexts within which a resource access can be executed. Each model is ascribed different priorities of which different policies were applied. The most important feature of this model is that the context representing plain resource access will not cause any race condition if all other threads are accessing from the same context. The result of our experiment shows that context-based synchronization performs better than Java given the same number of threads.