Workload Generation Techniques For Process-Aware Information Systems

Process Aware Information System (PAIS) is a kind of software which bases onprocess model, and is used to manage and execute business process. The businessprocess concerns people, applications and information resources. There is a trend thatbusiness models are decomposed from the implement of system engine, to comply withthe changing of business requirements. As workload of system engine, business modelof PAIS is one of the most important parts when doing system test. To do function andperformance test, we need quantity of workloads which have specific characteristics.This paper studies workload generation technologies of PAIS, and proposes how togenerate workloads with specific characteristics. The content of this paper contain threeparts.The test technologies for traditional information systems such as DBMS alwaysfocus on the data model, and few of them consider the business process issue; whilecurrent PAIS test technologies are always focus on process aspect, and do not take dataaspect into consideration. The isolation between process aspect and data aspect makesthe evaluation on PAIS to be neither comprehensive nor objective. For such reason, wepropose a PAIS workload generation framework, which can generate platformindependent model (PIM) for PAIS with model driven architecture (MDA). Thisframework considers process aspect and data aspect at the same time, and abstractscharacteristics of PAIS models to controllable parameters, to generate desired workloads.PAIS workload generation framework do not restrict the modeling language.This paperuse colored Petri net (CPN).When run-time data are given, there is a mapping between the construction-timeprocess model and the run-time instant for PAIS. Task expected execution number oftimes is a key metric for this mapping, which is important in PAIS workload generation.Existing solutions for such problem have their limitations. To resolve this problem forother general situations, we propose following approaches: token flow balance approach;divide and conquer approach; approach of referencing control-flow patterns. All theseapproaches can be used together under CPN model.ETL workflow is a typical application of PAIS used to integrate and transform databetween heterogeneous systems. Based on the ETL workflow modeled with CPN, this paper proposes approach to generate workload for ETL workflow. The generation ofcontrol-flow is studied at first. And then with the analysis of basic ETL activities,approach to generate data model based on attribute number constraints between datasets is proposed. At last based on the method of reverse query process and symbolic test,approach on how to generate semantic aware test data sets is proposed.PAIS workload generation techniques are studies in this paper, and the proposedworkload generation approaches are open, widely applicable and characteristiccontrollable, which can be used to evaluate the functionality and performance of systemengines and related algorithms. And also the completeness of proposed process modelgeneration algorithms are proved and the generation algorithms are validated byexperiments.