Measuring true performance of the work function algorithm for solving the on-line k-server problem

This paper deals with the work function algorithm (WFA) for solving the on-line k-server problem. The paper is concerned with assessing actual performance of the WFA in terms of its serving costs. First, an efficient implementation of the WFA is briefly described. Next, some experiments are presented, where the performance of the implemented WFA has been measured on very large problem instances. Thereby, the problem instances have been selected from some frequently studied classes where the most accurate theoretical estimates of performance are available. Finally, the measured performance of the WFA is compared with the corresponding theoretical estimates and with some other algorithms.