
PEPA examples
 Susanna Donatelli, Marina Ribaudo and Jane Hillston included an example in the paper A comparison of Performance Evaluation
Process Algebra and generalised stochastic Petri nets of a multiprocessor system, bus and memory modules.
 A small model of a location tracking system based on active badges was made by Stephen Gilmore, Jane Hillston and Graham Clark. The paper Specifying performance measures for PEPA appeared in the proceedings of Fifth International AMAST Workshop on RealTime and
Probabilistic Systems.
 A small model of the alternating bit protocol was developed by James Edwards. The paper Process algebras for
protocol validation and analysis appeared in the proceedings of the PREP 2001 conference in Keele.
 Two mediumsized PEPA examples are here. They model a PCLAN as described in lecture notes from the University of Edinburgh lecture course
`Modelling and Simulation''. The lecture notes explain how certain performance measures are calculated.
 A larger model of an industrial production cell was made by Robert Holton. The paper A PEPA specification of an industrial
production cell appeared in the Computer Journal.
 Jane Hillston, Marina Ribaudo and Stephen Gilmore made a model of the TOMP machine protocol. This model can be scaled very simply by varying the
number of copies of number of copies of the processes which are competing for memory access. The description of the model and timings for runs of the
PEPA Workbench on a 500MHz Pentium III with 128Kb of memory appear in the paper An efficient algorithm for aggregating
PEPA models which is to appear in IEEE Transactions on Software Engineering. Here is a collection of models of the TOMP machine with increasing
numbers of copies of components.
 PEPA model of TOMP (1,1,1) (72 states)
 PEPA model of TOMP (2,1,1) (176 states)
 PEPA model of TOMP (2,2,1) (416 states)
 PEPA model of TOMP (2,2,2) (960 states)
 PEPA model of TOMP (3,2,2) (2,176 states)
 PEPA model of TOMP (3,3,2) (4,864 states)
 PEPA model of TOMP (3,3,3) (10,752 states)
 PEPA model of TOMP (4,3,3) (23,552 states)
 PEPA model of TOMP (4,4,3) (51,200 states)
 PEPA model of TOMP (4,4,4) (110,592 states)
 Jeremy Bradley and Stephen Gilmore made a model of a web server which can be scaled by varying the number of servers, number of readers, number of
writers or buffer size. This can be used to generate systems with very large state spaces. The model is studied in the paper ``Derivation of
Passagetime Densities in PEPA models using ipc: The Imperial PEPA Compiler".
 PEPA model of WS (3,3,2,2) (1,376 states)
 PEPA model of WS (4,3,3,3) (21,248 states)
 PEPA model of WS (5,4,3,3) (69,440 states)
 PEPA model of WS (6,5,3,3) (211,968 states)
 PEPA model of WS (6,5,4,4) (1,369,728 states)
 PEPA model of WS (7,6,5,5) (26,651,520 states)
 PEPA model of WS (8,7,5,5) (> 26,651,520 states)
 PEPA model of WS (9,8,6,6) (>> 26,651,520 states)
 WS system generator (ML application)
 The mobile agent system is a very simple PEPA net which models a mobile software agent which moves from one host to another harvesting information
which it dumps to a central analysis point. The example is discussed in the paper PEPA nets: a structured performance modelling
formalism which appeared in the proceedings of Tools 2002.
 The concrete syntax for PEPA supported by the Möbius modelling platform differs slightly from the
syntax supported by other PEPA tools. Below is the tiny example in the Möbius syntax.
