Dynamic scheduling in grid systems

Grid computing is a relatively new distributed computing paradigm that is rapidly gaining importance. It offers a solution to the increasing demand of extensive computational and storage power, without requiring extraordinary investments in hardware infrastructure. However, in many cases the use of grids as such is insufficient without further optimizations, such as scheduling mechanisms for efficient assignment of tasks to available resources. This PhD research is dedicated to dynamic scheduling, which differs from static scheduling in that it allows for re-evaluation of already taken assignment decisions during task execution. Dynamically scheduled systems are more flexible, fault-tolerant and QoS-oriented, but also more complex and demanding than their static equivalents. Generic network-aware scheduling algorithms will be developed for grid systems with dynamically changing resource and task properties. Moreover, calibration and evaluation of the developed methods will be performed in a simulation environment and in several existing grid systems. Keywords— Grid computing, optimization, dynamic scheduling, QoS (Quality of Service), simulation, software development