|M.Sc Student||Katz Aric|
|Subject||A Distributed Simulation for Better Analyzing Hospital's|
|Department||Department of Industrial Engineering and Management||Supervisor||Professor Avraham Shtub|
|Full Thesis text - in Hebrew|
Advances in software and hardware development made large and complex simulation models feasible (for building), yet running them on a single computer remains a problem.
Prior studies show an exponential runtime increase when adding new models to a distributed simulation environment.
Although this study is focusing on hospital’s processes and operations, this study shows relevance to all complex simulation models (supply-chains, healthcare industry, semi-conductor fabs).
Part of the covered literature review showed the interest healthcare professionals have regarding process improvements in the E.R. and O.R.’s. Most of the studies that were reviewed delivered local solutions to local impediments and were never tested globally (regarding the whole hospital). A global examination \ analysis may expose local improvements for being damaging to the global system (shifting a bottleneck downstream).
Large and complex simulation models can suffer from extensive runtimes leading to an inefficient analytic tool. An alternative for runtime reduction can be delivered by “distributed simulation”. Distributed simulation can require multiple computers (one computer for one or more distributed simulation models) lead to expensive solution.
When running in a distributed environment simulation models communicate with one another via message sending on the network.
Distributed simulation infrastructure solved issues regarding runtimes, yet introduced new synchronization overheads.
In this study three models were developed:
- A distributed simulation infrastructures for Arena simulation (runtime reduction).
- A combined data-mining based simulation (runtime reduction).
- A generic simulation model for the hospital’s imaging center.
The distributed infrastructure developed was based on H.L.A. (Higher Level Architecture) tailoring it for Arena simulation.
The combined data-mining infrastructure developed is a novel approach for running distributed simulation models on a single computer and reducing the overall runtime in a dramatic manner.
The hospital’s generic imaging-center simulation model was developed using a novel approach for self generating simulations and an external data model which is based on organizational online data.
Results show that distributed simulation models deliver runtime reduction for large and complex simulations yet mid size simulations can show an increase in runtime when distributed.
Results from the combined data-mining infrastructure show runtime reduction for all complexities of simulation models.
The contributions of this study can be summarized to:
- A novel infrastructure for distributing complex simulation models on Arena.
- A novel infrastructure based on data-mining models for running distributed simulation models on a single computer.
- A generic self generated imaging-center’s simulation model based on external data-mining models.