## Timed Net Evaluation Tool (TimeNET)

** Category** Cross-Omics>Agent-Based Modeling/Simulation/Tools

** Abstract** TimeNET is a software tool for modeling and performability evaluation using stochastic Petri nets.

Its main characteristic is the evaluation of models with non-exponentially distributed firing delays.

TimeNET has been successfully applied during several modeling and performance evaluation projects and there are several hundred installations in universities and other organizations worldwide.

TimeNET Supported Net Classes and Analysis Methods --

The classical main model class of TimeNET is the extended Deterministic and Stochastic Petri Nets (eDSPNs).

Firing delays of transitions can either be zero (immediate), exponentially distributed, deterministic, or belong to a class of general distributions called expolynomial in an eDSPN.

Such a distribution function can be piecewise defined by exponential polynomials and has finite support. It can even contain jumps, making it possible to mix discrete and continuous components.

Many known distributions (uniform, triangular, truncated exponential, finite discrete) belong to this class.

Under the restriction that all transitions with non-exponentially distributed firing times are mutually exclusive, stationary numerical analysis is possible.

If the non-exponentially timed transitions are restricted to have deterministic firing times, transient numerical analysis is also provided.

For the case of concurrently enabled deterministically timed transitions, an approximation component based on a generalized phase type distribution has been implemented.

If there are only immediate and exponentially timed transitions, the model is a Generalized Stochastic Petri Net (GSPN) and standard algorithms for steady-state and transient numerical evaluation based on an isomorphic Markov chain are applicable.

The tool also comprises a simulation component for eDSPN models, which is Not subject to the restriction of only one enabled non-Markovian transition per marking. Steady-state and transient simulation algorithms are available.

Results can be obtained faster by parallel replications, using control variates, or with the RESTART method in the presence of rare events.

During the simulation run, intermediate results of the performance measures are displayed graphically together with the confidence intervals.

Simple stochastic Petri nets in TimeNET can either be interpreted in continuous time as an eDSPN, or as a Discrete Deterministic and Stochastic Petri Net (DDSPN).

DDSPNs allow geometric distributions, deterministic times and discrete phase type distributions as delays. Steady-state and transient numerical analysis as well as efficient parallel simulation are available.

Variable-free Colored Petri Nets (VfCPNs) -

Variable-free Colored Petri Nets (VfCPNs) represent another model class of TimeNET.

They were initially developed for the modeling and evaluation of manufacturing systems but are Not restricted to this area as shown in other applications.

Firing delays of transitions have the same range as in eDSPNs.

Numerical steady-state analysis, an iterative approximation method, and standard simulation are available for VfCPNs. A modeled (production) system can be directly controlled with this tool.

Stochastic Colored Petri Nets (SCPNs) -

Stochastic Colored Petri Nets (SCPNs) have been added recently during the ongoing transition to TimeNET version 4.0.

Due to the inherent complexity of the considered models, a requirement of only one non-exponential transition per marking is too restrictive.

Thus a standard discrete-event simulation has been implemented for the performance evaluation of SCPN models.

In addition, a distributed simulation method has been developed and implemented for this model class, which allows the efficient simulation of complex models on a cluster of workstations.

Both simulation programs automatically generate a set of C++ code fragments from an SCPN model, which implements the core functions of transitions and other model elements.

The results are compiled and linked with a run-time kernel of the simulation.

A model-specific and very efficient simulation program is thus obtained, which is then used either for a fast simulation in background mode or the token game in a single-step mode that cooperates with the Graphical User Interface (GUI).

Performance measures can be analyzed during run time with the graphical front end.

Note: The token game, i.e. an interactive simulation of the behavior of a model, is available for all mentioned model classes in TimeNET.

A Model-Class Generic Graphical User Interface --

The graphical user interface for TimeNET 4.0 has been completely rewritten in JAVA, and can therefore be run on both UNIX- and Windows-based environments.

The new GUI retains the advantages of the former one, especially in being generic in the sense that any graph-like modeling formalism can be easily integrated without much programming effort.

Nodes can be hierarchically refined by corresponding sub-models.

The GUI is Not restricted to Petri nets, and is already being used by other tools than TimeNET.

As a stand-alone program it has been named ‘PENG’, which is short for Platform-independent Editor for Net Graphs.

Model classes are described in an XML schema file, which defines the elements of the model. Node objects, connectors and miscellaneous others are possible elements.

For each node and arc type of the model the corresponding attributes and the graphical appearance is specified. The shape of each node and arc is defined using a set of primitives, and may depend on the attribute values of the object.

Actual models are stored in an XML file, consistent with the model class definition.

Program modules can be added to the tool to implement model-specific algorithms. A module can select its applicable net classes and extend the menu structure by adding new algorithms.

All currently available and future extensions of net classes and their corresponding analysis algorithms are readily integrated with the same “look-and-feel” for the user.

*System Requirements*

TimeNET 4.0 is currently supported on:

- 1) Windows
- 2) Linux on personal computers (recommended distribution: Debian 3.1)
- 3) SunOS 5.6 - 5.9 (Solaris) on Sun workstations.

*Manufacturer*

- Technical University of Ilmenau
- Systems and Software Engineering
- PF 100 565
- D-98684 Ilmenau, Germany

** Manufacturer Web Site**
Timed Net Evaluation Tool (TimeNET)

** Price** Contact manufacturer.

** G6G Abstract Number** 20612

** G6G Manufacturer Number** 104212