Scheduling and performance prediction of I/O intensive tasks in parallel time-sharing systems

Systems for the archival and retrieval of images are used in many areas. A search for objects and other important images components requires dynamic feature extraction. The related computational demands require utilisation of powerful parallel architectures. Therefore we developed a cluster platform, which supports the implementation of this novel retrieval approach in existing systems. In previous work we discussed the workload balancing across the cluster and presented a model to describe redistribution strategies. In this work we show how to incorporate the time-sharing mode into the model to reduce the response time of the system.