A Study of Performance Variations of Specialized Distributed Computing Systems Using Multiple Parallel Task Queues

Several variants of specialized distributed computing system models are considered, each containing one, two, three, or four parallel task queues. Such systems are characterized by continuous, non-stationary input streams of various types of tasks. The input tasks are parallelized into an a priori unknown number of data-independent subtasks. When a subtask is executed, all remaining branches of that task are deleted. The goal of this work was to study the behavior of such systems in homogeneous and heterogeneous environments when organized with multiple parallel queues. Experiments were conducted, and changes in the efficiency and performance of such systems with an increasing number of queues were examined. Research has shown that the use of multiple queues allows for increased performance without increasing computing resources.