See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/228719712

Developing a Meta-Methodology Supporting the Application of Parallel

Simulation

Article · January 2006

CITATIONS READS

6 1,515

2 authors:

Laszlo Muka Gábor Lencse

Széchenyi István University, Gyor Széchenyi István University, Gyor
23 PUBLICATIONS 85 CITATIONS 99 PUBLICATIONS 626 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

IPv6 Transition Technologies View project

MPT Network Layer Multipath Library View project

All content following this page was uploaded by Gábor Lencse on 30 May 2014.

The user has requested enhancement of the downloaded file.

DEVELOPING A META-METHODOLOGY SUPPORTING THE APPLICATION
OF PARALLEL SIMULATION
László Muka Gábor Lencse
Elassys Consulting Ltd. Department of Telecommunications
Bég utca 3-5. Budapest Univ. of Techn. and Econ.
H-1026 Budapest, Hungary Magyar tudósok körútja 2.
e-mail: [email protected] H-1117 Budapest, Hungary
e-mail: [email protected]

KEYWORDS lem with this approach is that there is a methodological gap

between SSM and methods to deal with simulation.
meta-methodology, parallel simulation, discrete-event- In our approach this methodological gap is eliminated by
simulation, organisational process, information and com- the development of modified conceptual models.
munication technology, Soft Systems Methodology, con-
ceptual model DEVELOPING MODIFIED CONCEPTUAL MODELS

ABSTRACT The Seven-stage Process of Traditional SSM

New concepts are described to SSM (Soft Systems Meth- Checkland’s SSM is an approach to apply systems-thinking
odology) conceptual models, which are tools for system to ill-defined problems in human activity systems. It is also
analysis supporting the application of simulation including described as a system-based problem-solving methodology
decisions about parallel simulation in an organisational starting with the unstructured problem situation. By the
environment. A meta-methodology facing with unstruc- outcome it is also defined as a learning system, a system for
tured problems in simulation projects and also supporting Operational Research or a method for information system
parallel simulation is formulated. analysis and design (Curtis 1989).
Stages of SSM are shown in Figure 1. The process of SSM
INTRODUCTION
seems to be linear: it is a sequence of well-defined stages
and there is a progression from one stage to the next in the
Simulation projects initiated to support Information and
methodology. Working with SSM is an iterative process,
Communication Technology (ICT) system design and
since it may be necessary to re-enter an earlier stage for re-
Business Process (BP) design in an organisation usually
execution.
begin with an unstructured problem situation, where fre-
quently there is on opinion that simulation takes a lot of
time and requires significant resources to be assigned with
the risk of getting no useful results.
In this paper we outline a meta-methodology addressing
these problems: we develop a soft approach to support
problem-structuring and effective goal definition to build
useful models and also increasing efficiency by precise
localization of systems to be modelled and by supporting
decisions on the use of parallel simulation helping in
speeding up the simulation.
In this paper we introduce new concepts to SSM (Soft Sys-
tems Methodology, Checkland 1985, 1989) conceptual
models then using the new concepts and a traditional six-
step process of simulation methodology we outline a simu-
lation meta-methodology.
Figure 1. The Seven-stage Process of SSM
Ideas about N&S (Necessary & Sufficient) conditions and
In Stage 1 and 2 there is a finding out about the unstruc-
“temporal relations” of conceptual models described by
tured problematical situation that is entering and expressing
Gregory (Gregory 1993) are used as starting point in our
the problem situation.
paper.
In Stage 3 relevant human activity systems are identified
In (Sierhuis and Selvin 1996, Sierhuis and Clancey 2002)
and using CATWOE analysis (Checkland 1989) root defi-
there is a description of a framework for collaborative
nitions of selected systems are formulated.
modelling and simulation using SSM and a set of four
methods to cover the modelling activities. The main prob-
In Stage 4 there is the conceptual model-building of rele- function, taken into account IT elements’ connected fea-
vant systems from the root definitions provided in Stage 3. ture.
Conceptual models are models of the views of what exist
In general, a function is performed if it is assigned to an
and not models of what exist in the real world. In a concep-
existing or a new organisational process and the necessary
tual model key activities of the system are taken into ac-
organisational resources (roles and responsibilities) are
count. A key activity itself generally represents a subsystem
assigned to the process. It means that using a PA is neces-
(Curtis 1989) that would carry the activity out, thus a hier-
sary only in special cases: in the case if the necessary proc-
archy of conceptual models can be defined when replacing
ess resources are not assigned in a PF and its PCs elements
a first-level conceptual model of a subsystem with its de-
(for example the necessary resources are assigned in a
tailed conceptual model.
shared way), or we want to examine the subsystem respon-
In Stages 4 - 7 there is a comparison with the real world to sible for the resource assignment.
define necessary and feasible changes and to define actions
to implement changes. In the case of an information system design agents can also
be IT-type elements, which are software and hardware re-
In the following points we harden up the methodology
sources.
(Jackson and Keys 1984) by introducing new concepts into
the conceptual models.
Now let us see a short example. Figure 2. shows a concep-
tual model of a Customer Request Processing System. After
Function elements in Conceptual Models
receiving the customer request by PF1 its processing is per-
In this paper we focus on the design of information systems formed by PF2, using information obtained by PC1 from
in an organisation therefore we may suppose that a key CRM (Customer Relationship Management) database. Cus-
activity is performed in general by an OP (Organisational tomer request is scheduled by PF3 (service activity assigned
Process) function or by an ICT system function. In other to customer request) using schedule information obtained
words it may be said that any function in the organisation by PC2 from service department, which is in another sys-
can be performed by some relevant organisational process tem. Answering the request is performed by subsystem PF4.
(P subsystem) with its human resources or by some relevant
IT subsystem with its technical resources.
PA1 and PA2 are agent elements guaranteeing resources for
Thus the subsystem elements in our conceptual models can functions in PF2 and PF3 to be performed.
be P-type or IT-type; depending on they represent OP or
ICT system function.
In our approach, an important feature of IT elements (ac-
cording to the traditional approach of SSM) is that any IT
element in the model should be connected to a minimum of
one P element in order to have its human resource connec-
tion. We may look at the conceptual model as a directed
graph CM(N;E), where N is the set of nodes containing P-
type or IT-type elements, E is the set of directed edges. In
order to define the connected feature of IT elements we
introduce a logical variable CON to describe that nodes x
and y of graph CM are connected:
⎧= 1, if ( x; y) ∈ E or ( y; x ) ∈ E
CON ⎨
⎩= 0, otherwise
where x ∈ P ∪ IT and y ∈ P ∪ IT

Now it may be said about IT elements:

∀ITi (i=1;2;…;I) ∃j (j=1;2;…;J) Figure 2. Conceptual Model with N&S Conditions after
(where I is the number of elements in the set IT, J is the Identifying PA, PF and PC Elements
number of elements in the set P)
In Figure 3. there can be seen the model of the same system
CON(ITi;Pj)=1 with one agent element PA1. It was decided not to use PA2
To describe the set of N&S conditions (Gregory 1993) we because in PF3 and in PC2 there is a sufficient assignment
define three element types F, C and A, It means that there of resources and we do not want to examine the resource
can be PF, PC, PA, ITF, ITC and ITA elements. PF is an assignment subsystem.
element performing basic function in the system; PC is pro- We express N&S conditions in symbols for PF2 and PF3:
viding conditional function necessary to perform basic
function while PA is an agent element ensuring the suffi- PF2 ⇔ (PF1 ∧ PA1 ∧ ePC1)
ciency for the basic function to be completed. ITF, ITC and PF3 ⇔ (PF2 ∧ ePC2)
ITA also perform subsequently basic, conditional and agent
In Figure 3. elements PC1 and PC2 are expanded (Check- A conceptual model’s virtual time is a time sequence as-
land 1985). PC1 contains subsystems ITC1.1 (the CRM signed to a conceptual model by giving time labels to ele-
function subsystem) operated by PC1.1. In PC2 there are ments. Time labels T(i) and T(i-1) have the meaning that a
subsystems ITC2.1 and PC2.1. where ITC2.1 can be an intra- function with time label T(i-1) performed earlier than a func-
net system function and PC2.1 a function to provide Service tion with time label T(i). (See in Figure 4.) There is nothing
Department’s scheduling information obtained using intra- said about the measure ∆T= T(i)- T(i-1). (To give an estimate
net function. The operating subsystem of intranet here is of ∆T, simulation method can be applied.)
not examined. In Figure 4. we show two conceptual models CM1 and
The conceptual model CM in Figure 3. can be described as CM2, where CM1 may be the Customer Request Processing
directed graph CM(N;E;TR) where TR is the set of tran- System from our previous example and CM2 system per-
sient edges. Transient edges connect elements in different forming services (Service Department).
conceptual models. (A conceptual model we got from an CM1 and CM2 are connected with request and answer con-
expanded element is also defined to be a different one.) In nections (RCM2-RCM1, ACM2-ACM1) which may be
Figure 3. elements PC1 and PC2 are expanded. They con- described as graphs’ transient edges
tain P-type and IT-type elements in different configura- (CM1.PFi;CM2.ePCx.PCx(1)),
tions. (CM2.PFu;CM1.ePCv.ITCv(1))
RC (Request from Customer) and AC (Answer to Cus-
tomer) are entry and exit edges of graph CM1.

Figure 3. Conceptual Model with Expanded Elements

The expanded elements ePC1 and ePC2 are also conceptual

models described by directed graphs CM.ePC1 and
CM.ePC2 (CM denotes the original conceptual model).
Edge (PC1.1;ITC1.1) in graph CM.ePC1 represents an op-
erator-type connection while edge (ITC2.1;PC2.1) in graph
CM.ePC1.2 shows a provider-type connection. An operator
P element is responsible for a function of an IT element,
while a provider P element is responsible for a function
using an IT element. The transient edges
(CM.ePC1.ITC1.1;CM.PF2), (CM.ePC2.PC2.1;CM.PF3) Figure 4. Synchronising Conceptual Models CM1 and CM2
connect elements of expanded subsystems to elements in through Conditional Elements
conceptual model CM.
We remark that IT and P elements in expanded subsystems
Virtual Time and Synchronisation in Conceptual Mod- have the same time label. (T( ) denotes a time label which is
els not significant in our analysis.)
Through transient edges and conditional elements (PCx and
Introducing time into the conceptual models can be done by
PCv) virtual times of conceptual models are synchronised.
assigning time label to elements. Giving time label T to an
After synchronisation of CM1 and CM2 we have the next
element has the meaning that the event of a ‘function is
relations:
performed’ takes place at T.
T(s-1)=T(k-2) and T(s-1)<T(k-1) SM is also an iterative-type methodology which is applica-
T(m )=T(t-2) and T(m)<T(t-1) ble for both P and IT elements. In point SM1 we explicitly
took into account a preliminary model design, which typi-
On the bases of synchronisation a decomposition of execu- cally takes place only implicitly.
tion time of functions can be made. The decision about parallel simulation usually is made in
step SM3 or SM4.
CONSIDERATIONS ABOUT APPLICATION OF
PARALLEL SIMULATION Outlining the meta-methodology with support for paral-
lel simulation
Note: We have not used any constraints on the type of
simulation (continuous, discrete, time-driven, event-driven, Now we outline a meta-methodology (MM) applying the
etc.) therefore our results may be used to the application of new concepts concerning conceptual models introduced in
Event-driven Discrete-Event Simulation (DES) which is in this paper, using the classic SSM together with SSM with
the focus of our interest. modified conceptual models and SM described in previous
In the case of information system design after the IT and P point.
function analysis, assigning virtual time and synchronisa-
tion of conceptual models, we can have a critical set of ele- The phases MM1-MM4 basically follow the progress of
ments to be simulated. SM but in MM2-MM3 there is a soft systems type progress
The critical set may be an interconnected set of IT and P also. In every phase classic SSM is applied if we are facing
elements but practically it is a set of at least one IT element an unstructured problem and modified conceptual models
connected to one P element. This is the situation to consider are applied concerning questions of simulation.
parallel simulation. Methodology steps based on our new concepts are listed in
In the case of one IT and one P subsystem, depending on MM3.
the focus of the simulation there can be three basic parallel The phases of meta-methodology are:
simulation decisions: (1) detailed simulation of both IT and
P subsystems; (2) detailed simulation of IT system with MM1. Goal definition
simulated P as process environment; (3) detailed simulation MM2. Identification of a widened set of relevant systems
of P with simulated IT system as environment. The P and IT
parts can act as the two segments of parallel discrete event MM3. Development of conceptual models containing sys-
simulation. They can be executed parallel by two intercon- tems to be simulated
nected processors. For all the three situations the use of the ƒ Identify P and IT subsystems, and elements to
Statistical Synchronisation Method (Pongor 1992) can be N&S conditions
considered as an inter-processor synchronisation method if ƒ Define time relations in models, synchronise mod-
there is a relatively slow speed of changes in subsystems’ els, make time decomposition
states. In situation (3) the method of TFA (Traffic Flow ƒ Define critical P and IT elements to be simulated
Analysis) (Lencse 2001) may be appropriate. Methods for ƒ Make decisions on partitioning and grouping of P
the parallel execution of the Combined DES and TFA elements for parallel simulation
(Lencse 2004) can be found in (Lencse 2005). ƒ Make decisions on partitioning and grouping of IT
If a subsystem seems to be too complex to be simulated in elements for parallel simulation
one model a further partitioning by expansion of the ele- MM4. Support for implementation
ment can be considered.
Expanding a P element we may get a set of P and IT sub- MM1 Phase of defining goals (SM1) has great importance:
systems while an expanded IT may contain only IT subsys- this is the basis for effectiveness and efficiency. Goals for
tems representing a set of sub-functions of the element. simulation project should be got from the organisational
If we have a situation with more IT and P elements group- goals and objectives by the way of goal partitioning and
ing or integrating elements may be appropriate. linking to the processes to be simulated. Soft method
should be used for learning the situation and for defining
SIMULATION META-METHODOLOGY requirements for simulation models.
SUPPORTING PARALLEL SIMULATION Some fast full simulation cycles may be necessary to make
clear the objectives. In this phase methods like TFA may be
The six-step process of simulation analysis method useful. Preliminary design of simulation models may be
produced taking into account the principle of parsimony
In order to use in formulation of a meta-methodology, in (Pied 1991).
the next point we describe a classic SM (Simulation Meth- MM2 In this phase a widened set of relevant systems is
odology). identified: (SM2) systems from where data should be get
SM is a six-step process comprising: (SM1) Defining goals for simulation (to identify and analyse sources of data),
(including preliminary design of models); (SM2) Gathering systems for which simulation results may be interesting and
and analysing data; (SM3) Model design and model build- systems probably to be simulated, that is all systems possi-
ing (SM4) Performing simulation (with as-is, what-if bly influenced by the simulation project. During data analy-
analysis, model verification and validation); (SM5) Analys- sis (SM2) typical and critical data configurations should be
ing simulation results; (SM6) Supporting implementation. defined for the whole interval of simulation, or if possible
for a longer time. Identification of typical and critical data Gregory, F. 1993. Cause, Effect, Efficiency and Soft Systems Mod-
configurations should be done for all relevant systems. els J. Opl. Res. Soc. Vol. 44, No. 4.
Pidd, M. 1991. Computer simulation methods In Operations Re-
MM3 First, conceptual models to be simulated are selected search in Management, Edited by Littlechild, S., and Shutler. M.,
and developed then the new methodological elements are Prentice Hall, UK.
applied. Sierhuis, M., Selvin, A.M. 1996. Towards a Framework for Col-
In the selected models P and IT elements are identified by laborative Modeling and Simulation, Workshop on Strategies for
building up a map for the identified elements. Virtual time Collaborative Modeling & Simulation, CSCW ’96, Boston, MA
Sierhuis, M., Clancey, W.J. 2002. Modeling and Simulating Work
system is introduced into conceptual models and after se-
Practice: A Method for Work System Design IEEE Intelligent Sys-
lecting P and IT elements planned to be simulated precise tems, September/October 2002, Vol. 17, No. 5. pp.32-41.
time label values are assigned. Synchronisations of models Elassys Consulting Ltd. 2005 "Iminet Network Expert System"
are made through appropriate conditional elements. On the http://www.elassys.hu
bases of synchronisation a decomposition of execution time Lencse, G. 2001. Traffic-Flow Analysis for Fast Performance
of functions can be made. Now we may have a critical set Estimation of Communication Systems Journal of Computing and
of P and IT elements to be simulated. Information Technology 9, No. 1, pp. 15-27.
Thinking in parallel simulation we make decisions about Lencse, G. 2004. Combination and Interworking of Traffic-Flow
further partitioning or grouping elements: in case of too Analysis and Event-Driven Discrete Event Simulation Proceedings
large subsystems we may try to use expansion for partition- of the 2004 European Simulation and Modelling Conference
(ESM®'2004) (Paris, France, Oct. 25-27. 2004.) EUROSIS-ETI,
ing. The use of Statistical Synchronisation Method and
89-93.
TFA can be considered. Lencse, G. 2005. Speeding up the Performance Analysis of Com-
After making parallel-sequential decisions the traditional munication Systems Proceedings of the 2005 European Simulation
simulation is completed (SM3,4). At these points soft cy- and Modelling Conference (ESM®'2005) (Porto, Portugal, Oct.
cles may be necessary to define the what-if scenarios and 24-26.) EUROSIS-ETI, 329-333.
also for verification and validation of simulation models. Pongor, Gy. 1992. Statistical Synchronization: a Different Ap-
proach of Parallel Discrete Event Simulation Proceedings of the
MM4 Analysis of simulation results (SM5) may lead to
1992 European Simulation Symposium (ESS'92) (Nov. 5-8, 1992,
going back to earlier points for further analysis and simula- The Blockhaus, Dresden, Germany.) SCS Europe, 125-129.
tion or even if the results are satisfactory, the exact under-
standing may require more soft cycles. BIOGRAPHY
Support for implementation of results (SM6) may consist of
making correction plans. In making correction plans, fur- GÁBOR LENCSE received his M.Sc. in electrical engi-
ther soft cycles and simulation may be useful. neering and computer systems at the Technical University
of Budapest in 1994 and his Ph.D. in 2000. The area of his
CONCLUSIONS research is (parallel) discrete-event simulation methodol-
ogy. He is interested in the acceleration of the simulation of
We have developed new concepts to SSM to modify con- communication systems. Since 1997, he works for the Szé-
ceptual models: chenyi István University in Győr. He teaches computer
- we have introduced a system of IT and P elements to help networks and networking protocols. Now, he is an Associ-
common analysis of ICT and BP systems taking into ac- ate Professor. He does R&D in the field of the simulation
count N&S conditions of performing functions, of communication systems for the Elassys Consulting Ltd.
- we have defined the virtual time and conceptual model since 1998.
synchronisation concepts for compatibility with simulation Dr Lencse works part time at the Budapest University of
methods, Technology and Economics (the former Technical Univer-
- we have examined how the parallel simulation decision sity of Budapest). There he teaches digital design and com-
can be supported in conceptual model analysis. puter architectures.
Using the results in developing modified conceptual models
we have outlined a meta-methodology dealing with un- LÁSZLÓ MUKA graduated in electrical engineering at the
structured problems in a simulation project and also sup- Technical University of Lvov in 1976. He got his special
porting the application of parallel simulation. engineering degree in digital electronics at the Technical
University of Budapest in 1981, and became a university
REFERENCES level doctor in architectures of CAD systems in 1987. Dr
Muka finished an MBA at Brunel University of London in
Jackson, M.C., Keys, P. 1984. Towards a System of Systems Meth- 1996. Since 1996 he has been working in the area of simu-
odologies J. Opl. Res. Soc. Vol. 35, No. 6. lation modelling of telecommunication systems, including
Checkland, P. 1985. Achieving “Desirable ad Feasible” Change: human subsystems.
An Application of Soft Systems Methodology J. Opl. Res. Soc. Vol. He is a regular invited lecturer in the topics of application
36, No. 9. of computer simulation for performance analysis of tele-
Checkland, P. 1989. Soft systems methodology In Rational Analy- communication systems, at the Széchenyi István University
sis for a Problematic World, Edited by J. Rosenhead, John Wiley
of Győr.
& Sons Ltd
Curtis, G. 1989. Business Information Systems Addison-Wesley,
Wokingham, UK.

View publication stats