Cost Model for the SAP Gamification Platform

S. Brunstein
Master Thesis
MT1302 (June, 2013)
Supervised by
o. Univ.-Prof. Dr. Michael Schrefl
Instructed by
Dr. Stefan Berger
Accomplished at
University Linz, Institute of Business Informatics - Data & Knowledge Engineering

Abstract (English)

In this master's thesis, a generic platform to support gamification of applications is the object of study. This Gamification Platform was recently invented and implements a new type of architecture. The performance of the platform is unknown since no prior knowledge exists for this complex type of architecture. Conducted experiments have shown that the response time and throughput vary extremely, which has to be analyzed and described more precisely with a cost model.

To support decision-making and be able to calculate costs of a gamification concept beforehand, performance and throughput of the Gamification Platform should be known prior to its implementation. Additionally, it has to be decided whether the applications and the platform should prefer to communicate in synchronous or asynchronous mode. The synchronous mode consumes less memory, whereas the asynchronous mode operates faster. A cost model allows the pre-calculation of performance and throughput for the generic Gamification Platform.

First, a queueing network model is built based on the architecture of the platform, which estimates the performance. Subsequently, cost factors are researched and measured in three categories: Users, rules, and infrastructure. The influence of each significant cost factor on performance is modeled with polynomials.

The accuracy of the presented cost model in synchronous mode has a Mean Magnitude of Relative Error of 0.15, the asynchronous cost model's Mean Magnitude of Relative Error is 0.115. For a real use case the Magnitude of Relative Error are 0.436 in synchronous mode and 0.148 in asynchronous mode. Errors due to exclusion of factors and simplifying assumptions will need to be overcome in future research.

Even though the Gamification Platform is generic, the presented cost model only covers the most relevant parts. However, even this cost model is already complex, and, more importantly, delivers accurate estimations of the platform's performance. Further need for research is emphasized to include more cost factors as well as their effects on each other, and to support the automation of model usage.