System Software

The paper discusses the principles of redundancy classification for the design of fault tolerant computer systems. The basic functions of classification: definitive, characteristic and predictive are presented. It is shown that the proposed classification of redundancy possesses substantial predictive power. The proposed classification is suitable for the analysis of roles of hardware and software to achieve system fault tolerance

Most of the contemporary research in wireless networks is primarily based on simulations or in-house small scale experimental setups that are highly customized for the experiment and hence difficult to re-use. Although, this may be useful for smaller experiments, it is very difficult to replicate the process without writing extensive scripts for controlling and collating the results of the experiment which may take up more time than the actual experiment in some cases. The main hindrance to facilitating experiments is the lack of a flexible framework that will allow researchers to conduct several different experiments with ease and to be able to repeat them as often as necessary for statistical consistency. The above considerations motivated the ORBIT testbed project, which is a multi-user experimental facility to support research on nextgeneration wireless networks. In this paper, we describe the software architectural framework to facilitate repeatable wireless experiments that provides essential services to choreograph experiments as well as automates the routine tasks of measurement collection, thereby allowing researchers to focus mainly on algorithms and data analysis. In particular, we address key considerations to be taken into account such as capturing all experiment parameters to facilitate repeated experimentation as well as a framework for data collection. We also present use cases to demonstrate the flexibility of the architecture to perform different types of experiments including mobility emulation along with key results and observations.