Modeling the Performance and Scalability of a SAP ERP System using an Evolutionary Algorithm

abstract

Simulating the performance behavior of complex software systems, like Enterprise Resource Planning (ERP) systems, is a hard task due to the high number of system components when using a white box simulation approach. This paper utilizes a black box approach for establishing a simulation model for SAP ERP systems on the basis of real world performance data, which is gathered by using a synthetic benchmark. In this paper we introduce the benchmark, called Zachmanntest, and demonstrate that by using an evolutionary algorithm basing on the results of the Zachmanntest, the exact performance behavior of the ERP system can be modeled. Our work provides insights on how the algorithm is parameterized e.g. for the mutation and crossover probability, to receive optimal results. Furthermore we show that the evolutionary algorithm models the performance and scalability of an ERP system with an error less than 3.2%. With this approach we are able to build simulation models representing the exact performance behavior of a SAP ERP system with much less effort than required when using a white box simulation approach.