Evaluating a two dimensional box packing algorithm on batch processing cluster job scheduling problem

A batch processing cluster job scheduling problem can be viewed as a problem of packing a set of two dimension boxes into a large rectangular container. Each job can be viewed as a box. The job runtime can be viewed as the height of the box while the job processor requirement can be viewed as the width of the box. This work aims to apply a two dimensional box packing algorithm on batch processing cluster job scheduling problems. All experiments are conducted using an event-driven simulator. A real job trace collected from the computer center is used as the simulator input. The experimental results show that the two dimensional box packing algorithm favors large jobs in terms of the number of processors.