Distributed Augmented Reality Systems: How Much Performance is Enough?

Abstract

This paper presents architectural performance requirements for a framework dedicated to Augmented Reality (AR) applications. This work has been made as a starting point to evaluation of a framework named ARCS (Augmented Reality Component Systems) allowing rapid-prototyping, development, debugging and monitoring of AR applications. The article first introduces the need of such a dedicated framework and the importance of distribution for AR. It also identifies main features that should be offered by an AR framework and key performance criteria to be satisfied. These are partially inspired by the SPE (Software Performance Engineering) distributed architecture evaluation approach. These criteria are application specific. Thus, we first classify different distributed AR applications in scenarios and extract relevant criteria for each. We also provide a custom assessment protocol inspired by SPE evaluation methodology and based on previously identified criteria. This assessment is application independent as it is based on generic design patterns and most likely scenarios of distribution. Keeping in mind identified criteria and handling existing framework's constraints, a reliable generic AR framework should satisfy all criteria for each application class. Assessment protocol results for each criterion are compared to the typical features and performance that should be offered by an AR system.