Consistent, interactive steering of distributed computations : algorithms and implementation
MetadataShow full item record
Interactive computational steering provides users with the opportunity to tackle new problems in a way that helps them to learn about the computation in a highly engaging, interactive, visual environment. Causal consistency is an important feature of interactive steering of distributed computations, as it is often required to maintain the correctness of the computation. However, due to the asynchronous nature of distributed computations, it is difficult to coordinate steering changes across processes to guarantee that the changes are applied consistently at all processes. This thesis introduces a transaction-based computation model for distributed computation. This abstract model not only gives users a simple and high-level view of distributed computation, but also simplifies reasoning consistency problem by reducing the amount of information to be handled. Furthermore, this work investigates two approaches for achieving consistent steering: conservative steering and optimistic steering. The performance of conservative and optimistic steering approaches is evaluated in term of perturbation and lag. Our experiments show that when the percentages of consistency on the first attempt are large enough and the size of checkpoint is not too large, the optimistic approach will achieve better performance; otherwise, the conservative approach will be better.