“…However, makespan minimization is not suitable for large-scale distributed platforms as acquiring all various parameters (e.g., workload of individual tasks and the resource availability of all nodes) might take long time, which is also error-prone. Considering the heterogeneity of computing nodes, many other task scheduling solutions [13,14,15,16,17] are designed based on a fairness index criterion of nodes' workload levels. Unlike the traditional concept of fairness on user tasks' service level (e.g.…”