A load-balanced exact solver for computing the exact solutions of minimum k-center is described. To achieve the load balance on a dedicated multiprocessor system a new algorithm for parallel generation of a set of all k-combinations without repetitions is introduced and analysed. We demonstrate that the new algorithm can also be used in a resource competitive environment if used or supplemented with a simple adaptive job scheduler. The solver is tested by producing the benchmarks for the minimum k-center problem.
COBISS.SI-ID: 6954836
We present an attempt to improve job scheduling over heterogeneous GRID nodes by employing machine learning methods. Our aim is to provide a plugin which can be easily added to existing frameworks, thus avoiding significant modifications. We assume that existing scheduling algorithm in the framework should not be completely overridden, but rather modified only if there are chances that the modification will yield a better result. We focus on use of off-the-shelf machine learning methods in a black-box manner. The results show that improvements over the simple scheduling algorithms can be made.
COBISS.SI-ID: 0000000
This article focuses on mapping jobs to resources with use of off-the-shelf machine learning methods. In the article we focus on two sets of experiments, both constructed with a goal of congesting the system. In the first part, the machine learning methods are used as assistants for the base resource selection algorithms, while in the second part, they are used to directly select the resource for job execution. The results show that even with a such black-box approach of the off-the-shelf machine learning methods, we can achieve improvements or at least get comparable results.
COBISS.SI-ID: 0000000