A Novel Adaptive Support Vector Machine based Task Scheduling

Design and implementation of a cloud computing service for finite element analysis

“…Their results suggest that SVM scheduler can closely follow the performance of the best performing heuristics (Early First, Light Least) and soundly outperform Round Robin scheduler. Our work differs from [30] and other prior scheduling work such that we can split the FEA tasks further and choose among MP vs. MPI parallelization models before scheduling tasks onto machines. Therefore, we are neither bound to nor depend on the performance of other heuristics.…”

“…Chiang et al [22] address the Flexible Manufacturing Scheduling (FMS) problem with Critical Ratio-Based heuristics and genetic algorithm. Other algorithms called dispatching rules [25] include FIFO, Shortest Processing Time (SPT) [30], Critical Ratio (CR), EDD (Earliest Due Date), and Shortest Remaining Processing Time (SRPT). Park et al [30] proposed a multi-class SVM-based task scheduler for heterogeneous grids that maps queued tasks to machines based on their sizes and machines' loads as well as their machine computing powers to minimize the total completion time (makespan) of all tasks.…”

“…Other algorithms called dispatching rules [25] include FIFO, Shortest Processing Time (SPT) [30], Critical Ratio (CR), EDD (Earliest Due Date), and Shortest Remaining Processing Time (SRPT). Park et al [30] proposed a multi-class SVM-based task scheduler for heterogeneous grids that maps queued tasks to machines based on their sizes and machines' loads as well as their machine computing powers to minimize the total completion time (makespan) of all tasks. Their results suggest that SVM scheduler can closely follow the performance of the best performing heuristics (Early First, Light Least) and soundly outperform Round Robin scheduler.…”