=Paper=
{{Paper
|id=Vol-489/paper-11
|storemode=property
|title=MDPE Workbench - A Solution for Performance Related Decision Support
|pdfUrl=https://ceur-ws.org/Vol-489/paper11.pdf
|volume=Vol-489
|dblpUrl=https://dblp.org/rec/conf/bpm/FritzscheJCG09
}}
==MDPE Workbench - A Solution for Performance Related Decision Support==
https://ceur-ws.org/Vol-489/paper11.pdf
MDPE Workbench - A Solution for Performance
related Decision Support
Mathias Fritzsche1 , Jendrik Johannes2 , Sebastian Cech2 , Wasif Gilani1 ,
1
SAP Research CEC Belfast
United Kingdom
mathias.fritzsche@sap.com and wasif.gilani@sap.com
2
Technische Universität Dresden
Germany
jendrik.johannes@tu-dresden.de and sebastian.cech@tu-dresden.de
Abstract. Continuous optimization of business process execution in-
cluding its adaptation to changes within the market environment or the
company itself becomes increasingly important. Therefore, business per-
formance related decision support needs to be an integral part of Business
Process Management (BPM) Environments. We experienced that such
a support is especially useful when dealing with complex and resource
intensive business processes, such as business processes with layered use
of resources or with very complex workflows. Also the statistical distri-
bution of the history data or the plan data related to a process might be
a reason for such support. In this paper we present our generic Model-
Driven Performance Engineering (MDPE) Workbench, which seamlessly
integrates different kinds of existing Performance Analysis Tools, such as
discrete event simulation tools and optimization tools, with the design-
time and runtime of different BPM Environments. In more detail, the
MDPE Workbench allows a push-button performance related decision
support based on existing process models and process instance data with-
out requiring deep performance expertise of BPM users.
1 Introduction
Continuous optimization of business process execution, including its adaptation
to changes within the market environment or the company itself becomes in-
creasingly important. Therefore, business performance related decision support
needs to be an integral part of Business Process Management (BPM) environ-
ments. We experienced that such support is especially useful in cases of a high
degree of complexity in resource intensive processes, such as in the case of lay-
ered use of resources or complex workflows. Also the statistical distribution of
the history data or the plan data related to a process might be a reason for such
support.
We further experienced the need to provide support related to throughput
and utilization of resources. Additionally, predictions of gross execution time
of process instances are important in order to answer questions like “How will
2
the execution time of a Sales Ordering Process be affected by a certain planned
increase of the number of Sales Orders?”. Such predictions are supported by
discrete event simulation tools, such as AnyLogic [1]. Some of the existing BPM
anvironments already provide basic simulation capabilities. However, most of
them, such the EMC Documentum Process Suite [2], are limited when it comes
to sophisticated simulation features, such as resource sharing scenarios between
different departments.
Additional decision support need to be provided via the automated execution
of numerous what-if simulations for the purpose of business optimizations, for
instance in order decide at which point of time a certain business resource is
needed to meet processing targets and to optimize revenue. For optimization,
existing libraries and tools can be used, such as OptQuest [3].
Moreover, we experienced the need for analytical bottle-neck analysis. Es-
pecially if resources are used in a layered way, such as in the case of a Sales
Ordering Process where certain process steps can only be finalized if support
from a invoicing-department is provided which is itself dependent on the IT-
department. If the Sales Order Processing is not processed within targets, one
needs support to decide if the sales-department, Invoicing-department or IT-
department is the bottle-neck of the process. Such questions can be answered
with help of analytical bottle-neck analysers, such as the LQNSolver [4].
Model-Driven Performance Engineering (MDPE) provides a solution which
enables to extend existing BPM Environments with multi-paradigm decision
support by integrating existing Performance Analysis Tools like the AnyLogic
simulation tool, the OptQuest optimization tool and the LQNSolver for analytic
bottle-neck analysis.
The remaining part of the paper is structured as followed: The next section
describes one use case of MDPE which enables business performance related
decision support across multiple BPM Environments. In Section 3, an overview
on the architecture of MDPE is provided. Finally, in Section 4, we describe the
experiences we gained with our approach and follow-up steps.
2 Use Case
Real world business processes might be based on different BPM Environments,
for instance in the case of supply chains or processes or in the case a long-
running business process which can be extended for changing business conditions
with a Composite Application [5] having a comparable short lifecycle. In [6] we
provide one example for the last case. In this example, a manager of the Sales
and Distribution Organization of the Wine Seller wants to extend the existing
business process, Sales Order Processing in this case, which is provided by the
SAP Business Suite so that it is supported to add an extra free bottle of wine to
orders of those customers who notified a quality issue for their previous order.
This raised the need for an extended version of the Business Suite process. It was,
however, not desirable to change the business process directly in the Business
Suite because the application should be independent of the software vendor life
3
cycle. Therefore, a Composite Application has to be implemented. In our case, we
chose the NetWeaver BPM Environment [5] for this task. Other environments,
such as JCOM [7] or the EMC Documentum Process Suite [2] can be used for
this purpose as well.
However, business performance related decision support is needed for the
manager of the Sales and Distribution Organization to indicate how the planned
Composite Application will affect the end-to-end business process spanning Sales
Order Processing and the process extension. In order to enable such support,
we were required to extend the NetWeaver BPM Environment and BPM tooling
provided for the management of Business Suite processes, in our case a SAP
proprietary modelling tool and process monitoring tools such as the tooling
called Process Performance Manager (PPM) by IDS Scheer [8]. The following
section describes the MDPE Workbench a solution for this need.
3 MDPE Workbench Architecture
Figure 1 shows that the MDPE Workbench enables to interconnect arbitrary
BPM Environments with Performance Analysis Tools of choice which enables
multi-paradigm decision support.
!!!
!!!
5
!!!
!!!
"#
"#
5
"#
"#
Fig. 1. Architecture of the MDPE Workbench as Block Diagram [9]
The central part of the MDPE Workbench, the so called MDPE Core com-
bines Performance Parameter, such as how much time was consumed by a spe-
cific activity, with Process Models. Performance Parameters are either assumed
and/or planned values, or data which are approximated from historical process
instance data with help of a heuristic [10]. As central model within the MDPE
Core, a so called Tool-Independent Performance Models (TIPM) is used, which
associates behaviour information from the Process Models with its related re-
source consumption from the Performance Parameters. The TIPM is used as
input to create Tool-Specific Performance Models (TSPMs) which are contain-
ing input for a Performance Analysis Tool, such as the input files to execute a
discrete event simulation in AnyLogic [1].
4
This decomposed transformation approach via the TIPM has been choosen
in order to reduce the number of required model-to-model transformations as
we are required to interconnect n different BPM Environments with m kinds
of Performance Analysis Tools. Thus, the TIPM simply reduces the number of
required transformations from n ∗ m to n + m. Beside of capabilities to generate
input for Performance Analysis Tools, the MDPE Core enables to analyse the
Performance Analysis Results, for instance by checking if a prediction meets all
user provided targets (see Result Analyzer in Figure 1). Additionally, the MDPE
Core enables to trace Decision Support Results back with help of the Tracing
Controller [11].
The MDPE Workbench is configured via the MDPE Administration Tool in
order to specify which BPM Environment Adaper and Performance Analysis
Tool are active to interconnect BPM Environments and Performance Analysis
Tools. This configuration, including the paths of all Process Models, the TIPM
and all TSPMs, are strored within a so called megamodel [12] which is used as
input for the Transformation Controller to locate the required models when the
transformations between the BPM Environments and the Performance Analysis
Tools are executed.
Furthermore, different kinds of relationships between modelling artefacts can
be stored in the megamodel. For instance, the relationship between the different
models of the transformations provided by the Source- and Target- Adapters are
stored within this model. This information is accessed by the Tracing Controller
for the tracing of Performance Analysis Results back to the BPM Environment.
4 Experiences and Future Work
In this paper, the basic architecture of the MDPE Workbench has been de-
scribed. This tooling enables to interconnect arbitrary BPM Environments with
multi-paradigm decision support by integrating existing Performance Analysis
Tools. We successfully evaluated the MDPE Workbench by extending the JCOM
BPM Environment, the NetWeaver BPM Environment and a SAP proprietary
process modelling tools for SAP Business Suite processes with the AnyLogic
simulation tool, the OptQuest optimization tool and the LQNSolver bottle-neck
analyser.
The flexibility of the MDPE Workbench has been identified as especially
beneficial when we experimented with performance related decision support for
the extension of the Sales Ordering Process with a Composite Application for
the needs of a wine seller as described in Section 2. In that case, we had to extend
the NetWeaver BPM Environment and the BPM tooling for SAP Business Suite
processes at the same time.
As Process History Data we utilized assumed data in that case as we haven’t
integrated the relevant business process logs yet. We however experienced that
such integration is beneficial as we are already able to consume historical pro-
cess instance data from the business log of the JPASS BPM Environment as
this prevents users to manually annotate assumed Performance Parameters [10].
5
Therefore, as follow-up step we anticipate to integrate the business process log of
the NetWeaver BPM Environment and process logs for Business Suite processes,
such as the PPM tooling to further evaluate our tooling on real-world processes.
Moreover, we are currently employing an approach for user guided uncertainty
reduction for cases where the data quality of the logged data is low, for example
due to a high variance [10].
Disclaimer
The information in this document is proprietary to the MODELPLEX consor-
tium members SAP AG, and TU Dresden. The information in this document
is provided “as is”, and no guarantee or warranty is given that the information is
fit for any particular purpose. The above referenced consortium members shall
have no liability for damages of any kind including without limitation direct,
special, indirect, or consequential damages that may result from the use of these
materials subject to any liability which is mandatory due to applicable law.
Copyright 2008 by SAP Research and TU Dresden.
References
1. XJ Technologies: (AnyLogic — multi-paradigm simulation software) URL
http://www.xjtek.com/anylogic/.
2. Associates, B.S.: The BPMS Report: EMC Documentum Process Suite 6.0 (2007)
3. Ycesan, E., h. Chen, C., Snowdon, J.L., Charnes, J.M.: (Optimum-seeking simula-
tion in the design and control of manufacturing systems: Experience with optquest
for arena)
4. Franks, R.G.: Dissertation: Performance analysis of distributed server systems
(1999)
5. Snabe, J.H., Rosenber, A., Molle, C., Scavillo, M.: Business Process Management:
The SAP Roadmap (2008)
6. Fritzsche, M., Gilani, W., Fritzsche, C., Spence, I.T.A., Kilpatrick, P., Brown, T.J.:
Towards utilizing model-driven engineering of composite applications for business
performance analysis. In: ECMDA-FA, LNCS 5095, Springer (2008)
7. JCOM: http://www.jcom1.com/ (2008)
8. IDSScheer: http://www.ids-scheer.com (2009)
9. Knöpfel, A., Gröne, B., Tabeling, P.: Fundamental Modeling Concepts: Effective
Communication of IT Systems. John Wiley & Sons (2006)
10. Fritzsche, M., Kilian-Kehr, R., Gilani, W.: Guided uncertainty reduction in au-
tomatically generated business simulations. In: To appear in proceedings of the
IEEE SIMUL’09 conference. (2009)
11. Fritzsche, M., et al., J.J.: Systematic usage of embedded modelling languages in
model transformation chains. In: SLE’08, LNCS 5701, Springer (2009)
12. Fritzsche, M., Bruneliere, H., Vanhoof, B., Berbers, Y., Jouault, F., Gilani, W.:
Applying Megamodelling to Model Driven Performance Engineering. In: 16th IEEE
ECBS’09, IEEE (2009)