Please use this identifier to cite or link to this item: http://dx.doi.org/10.18419/opus-9613
|Title:||Workload mix definition for benchmarking BPMN 2.0 Workflow Management Systems|
|Abstract:||Nowadays, enterprises broadly use Workﬂow Management Systems (WfMSs) to design, deploy, execute, monitor and analyse their automated business processes. Through the years, WfMSs evolved into platforms that deliver complex service oriented applications. In this regard, they need to satisfy enterprise-grade performance requirements, such as dependability and scalability. With the ever-growing number of WfMSs that are currently available in the market, companies are called to choose which product is optimal for their requirements and business models. Benchmarking is an established practice used to compare alternative products and leverages the continuous improvement of technology by setting a clear target in measuring and assessing performance. In particular, for service oriented WfMSs there is not yet a widely accepted standard benchmark available, even if workﬂow modelling languages such as Web Services Business Process Execution Language (WS-BPEL) and Business Process Model and Notation 2.0 (BPMN 2.0) have been adopted as the de-facto standards. A possible explanation on this deﬁciency can be given by the inherent architectural complexity of WfMSs and the very large number of parameters aﬀecting their performance. However, the need for a standard benchmark for WfMSs is frequently aﬃrmed by the literature. The goal of the BenchFlow approach is to propose a framework towards the ﬁrst standard benchmark forassessing and comparing the performance of BPMN 2.0 WfMSs. To this end, the approach addresses a set of challenges spanning from logistic challenges, that are related to the collection of a representative set of usage scenarios,to technical challenges, that concern the speciﬁc characteristics of a WfMS. This work focuses on a subset of these challenges dealing with the definition of a representative set of process models and corresponding data that will be given as an input to the benchmark. This set of representative process models and corresponding data are referred to as the workload mix of the benchmark. More particularly, we ﬁrst prepare the theoretical background for deﬁning a representative workload mix. This is accomplished through identiﬁcation of the basic components of a workload model for WfMS benchmarks, as well as the investigation of the impact of the BPMN 2.0 language constructs to the WfMS’s performance, by means of introducing the ﬁrst BPMN 2.0 micro-benchmark. We proceed by collecting real-world process models for the identiﬁcation of a representative workload mix. Therefore, the collection is analysed with respect to its statistical characteristics and also with a novel algorithm that detects and extracts the reoccurring structural patterns of the collection.The extracted reoccurring structures are then used for generating synthetic process models that reﬂect the essence of the original collection.The introduced methods are brought together in a tool chain that supports the workload mix generation. As a ﬁnal step, we applied the proposed methods on a real-world case study, that bases on a collection of thousands of real-world process models and generates a representative workload mix to be used in a benchmark. The results show that the generated workload mix is successful in its application for stressing the WfMSs under test.|
|Appears in Collections:||05 Fakultät Informatik, Elektrotechnik und Informationstechnik|
Files in This Item:
|Dissertation_Marigianna_Skouradaki_WorkloadMix4WfMS.pdf||2,03 MB||Adobe PDF||View/Open|
Items in OPUS are protected by copyright, with all rights reserved, unless otherwise indicated.