Dynamic Resource-Critical Workflow Scheduling in Heterogeneous Environments Yili Gong 1⋆ , Marlon E. Pierce 2 , and Geoffrey C. Fox 3 1 Computer School, Wuhan University, Wuhan, HuBei, P.R.China 430079 yiligong@whu.edu.cn, 2 Community Grids Lab, Indiana University, Bloomington, IN 47404 mpierce@cs.indiana.edu 3 Community Grids Lab, Department of Computer Science, School of Informatics, Indiana University, Bloomington, IN 47404 gcf@indiana.edu Abstract. Effective workflow scheduling is a key but challenging issue in heterogeneous environments due to the heterogeneity and dynamism. Based on the observations that not all tasks can run on all resources and acquired data transferring and queuing for a resource can be concurrent, we propose a dynamic resource-critical workflow scheduling algorithm which take into consideration the environmental heterogeneity and dy- namism. We evaluate its performance by simulations and show that it outperforms another selected widely used approach. Key words: Dynamic Scheduling, Resource-Critical Scheduling, Work- flow, Heterogeneous Environments. 1 Introduction Heterogeneous distributed systems are widely deployed for executing compu- tation and/or data intensive parallel applications, especially scientific work- flows [1]. A workflow is a set of ordered tasks that are linked by logic or data dependencies and a workflow management system is employed to define, manage and execute these workflow applications [12]. The efficient execution of workflows in this kind of environments requires an effective scheduling strategy which de- cides when and which resources the tasks in a workflow should be submitted to and run on. The environment includes both heterogeneous resource and policy. The soft- ware installation and configuration on resources are different as well as their physical computing capabilities. On the other side, the administration policies, such as access control policies, are autonomous and diverse. The dynamism means that the resource status, e.g. load, waiting time in the queue, availability, etc., changes over time are often uncontrollable. Thus the environment requires ⋆ This work is supported by 973 Program (Grant No. 2005CB321807) and the National Natural Science Foundation of China (Grants No. 60773058 and No. 60672051).