We plan to create a device driver that helps manage a Linux Cluster maintaining high-availability and reliability in the cluster.

Complex systems with dependencies cannot be easily characterized neither by fault trees nor by RBD solutions. Hence we intend to use Markov models to perform the dependency analysis. We are working on design tools that will help us pin point problem areas on modeled systems and perform a cost based analysis to obtain a reliability estimation of the system.

Being able to perform statistical analysis of modeled systems can help us design more fault tolerant and reliable systems. Cost can be a prohibiting factor and hence being able to model based on cost considerations can be a handy tool for the system designer.

Benefits: A tool that will enable system designers and architects to estimate the target system reliability and perform cost based calculation while implementing an optimal system.

Distributed computing architectures can be difficult to model due to the complexities involved and the number of dependencies. There can be various problem domains under consideration and we would want to investigate appropriate reliability modeling techniques for our problem domains.

There are various reliability modeling tools that available, we plan to study them in detail and understand the strengths and limitations of the existing tools. This will enable us to design a better tool that can include features from the existing tools so as to avoid reinventing the wheel and thereby also create a tool that is a significant enhancement over the existing prototypes.

Benefits: A tool that will enable system designers and architects to estimate the target system reliability and evaluate potential points of failure in the system typically targeting distributed computing architectures.