pypepa
pypepa is a PEPA library and a toolset for Performance Evaluation Process Algebra (PEPA) by Jane Hillston. pyPEPA is not a fully PEPA compatible tool, it supports a limited (for now) PEPA syntax (we only allow <> operator in system equation), i.e. it does not suport hiding operator (e.g. P\{a,b,}), does not calculate passage time. pyPEPA also does not use Kronecker state space representation and Hillston's aggregation algorithms, so it can have worse performance than the PEPA Eclipse Plugin. All these features, plus more, are planned to be added in next versions. If you are willing to help, just email me or put a pull request.
Warning: pyPEPA is under development, this is a beta software
pyPEPA consist of three parts:
- libpepa - a library written in Python,
- pyPEPA - a command line tool for solving and graphing,
- distr/ - map reduce tools for solving large PEPA experiments.
More docs can be found on readthedocs.
News
(22.07.2013) Docs added, pypepa has now docs on rtd.org (18.07.2013) pypepa can now calculate utilisations of components' states, output argument works again (07.06.2013) Added support for defining rates as mathematical expressions, e.g. r=2*3+7*n;
Installation
Package
Using pip:
Manually:
- Clone the project
- Run install
From the source
For the current version I recommend installing in a virtualenv.
- Clone the project
- Make a virtualenv
- Install all requirements
Using pypepa
Basic arguments
Show help command:
Set logging level (the default is NONE):
Calculations
Calculate steady state for bank scenario. The putput is by default directed to your terminal.
Calculate actions' throughput:
You can calculate transient time proability for some number of time steps:
You can choose a solver by specifying --solver|-s {direct, sparse}. By defalt we use sparse solver with LIL matrix becuase it is faster and in overall matrices generated from PEPA models are sparse. There is also an insignificant difference in results.
pypepa allows you to visualise all PEPA components and the whole state space of a model by specifying -gd switch. The generated graphiz dot files are by deault saved in dots folder in the current directory. You can browse dot files with xdot, which you need to install first.
Finally pypepa can provide us with a tool for experimentation with rates and actions. Let's check how throughtput of askManager action changes when rateReset changes from 1 to 50 with step 1. The default result of this command will be a matplotlib graph. The format of -var is "vartype:varname:value range specifier:value range value". The one valid vartype for now is rate, for value range specifiers you can choose: range or list. For range you need to provide START, STOP, STEP, whereas for list a comma separated list of values. You can specify other output options with -f argument: graph, console, csv.
Formatting
You can specify formats of -st, -th and --varrate with a --format option. Currently we support CSV (although ; not comma delimited), console (the default) and graph (only for varrate experiments). Additionally you can specify -o|--output option with a file argument to specify where to save the CSV.
TODO
Functional
- Implement rate mathematical expressions with functional rates (DONE)
- Implement passage time analysis
- Implement hiding operator
- Implement 3d graphs and experiments (DONE)
- Implement Kronecker state space and aggregation
- Implement generalised communication PEPA genPEPA by Mirco Tribastone
- Add model manipulation language for reducers
- Add stochastic probes
- Add distributed version of BU algorithm
Non functional
- Optimise optimise optimise
Licence and credits
Copyright (c) Dariusz Dwornikowski and Poznan University of Technology. Distributed under the Apache Commons 2.0.