Layered queueing networks (LQN), as an extended queueing network, are used widely to evaluate many distributed systems which have a client-server architecture. However, LQN models also share the difficulty that convention queueing networks have in modelling state-based behavior, such as the decisions made in exception handling during resource allocations. In order to enhance the modelling power of LQN models to handle decisions, this paper defines four decision patterns: Abort, timeout, infinite-retries and finite-retries, which are commonly used in the exception handling during resource allocations. These four decision patterns are generalized to two cases: Timeout and retry decisions and implemented in the LQN simulation tool, LQSIM. The LQN input language was modified to allow these actions to be specified directly. The simulator was then verified by comparing its results from solving a model a small-scale web server model to results found from solving a Petri net model using GreatSPN. The results were quite similar to each other despite the extensive simplifications in the Petri net model required for solution.

Additional Metadata
Keywords Decisions, Exception handling, Layered queueing networks, Timeouts
Conference 48th Summer Computer Simulation Conference, SCSC 2016, Part of the 2016 Summer Simulation Multi-Conference, SummerSim 2016
Citation
Li, L. (Lianhau), & Franks, G. (2016). Modelling decisions in layered queueing networks. In Simulation Series (pp. 181–188).