Simulating layered queueing networks with passive resources

This paper describes an extension to Layered Queueing Networks (LQN), a form of an extended queueing network used to investigate performance problems, to model passive resources such as counting semaphores and buffers. Layered queueing networks can be constructed directly, or from UML design models which incorporate the MARTE profile, either directly or via the Core Scenario Model. Layered Queueing Networks scale well and can solve analytically systems with nested resource requests to active resources. However, passive resources cause problems which force the use of simulation. The layered queueing network simulator, lqsim, is also described here. Simulations are created by reading in an LQN model, constructing objects from pre-existing templates, then solving. The semaphore task extension was incorporated by modifying the existing template used to model multi-server tasks. Finally, the semaphore extension was used to solve a model of a building security system which has a pool of buffers to capture video images. The results here show that a lack of buffers is indeed a bottleneck, but other parts of the system ultimately limit the capacity of the system.