Table 2.
Characteristics of the real-word workflows.WorkflowNumber of jobsNumber of edgesAverage data size (MB)
Average job runtime (s)
Montage_1000100044853.21
11.36
Epigenomics_9979973228388.59
3858.67
CyberShake_100010003988102.29
22.71
Inspiral_1000100032468.90
227.25
*When calculating the number of edges, average data size and average task runtime, the pseudo entry/exit VX-11e and the related edges are included.Full-size tableTable optionsView in workspaceDownload as CSV
Fig. 7. The structure of evaluated workflows. (a) Montage. (b) Epigenomics. (c) CyberShake. (d) Inspiral.Figure optionsDownload full-size imageDownload as PowerPoint slide
Besides anaphase real world workflows, we also test our algorithm on workflows generated by a DAG generator tool, which was also used in [49]. The tool defines the DAG shape based on four parameters: width, regularity, density, and jumps. In our experiment, for random DAG generation, the parameters are set as width = [0.2, 0.4, 0.8], regularity = [0.2, 0.4, 0.8], density = [0.2, 0.4, 0.8] and jumps = [1, 2, 3]. With these parameters, one DAG with 1500 jobs is created by choosing the value for each parameter randomly from the parameter data set.
Characteristics of the real-word workflows.WorkflowNumber of jobsNumber of edgesAverage data size (MB)
Average job runtime (s)
Montage_1000100044853.21
11.36
Epigenomics_9979973228388.59
3858.67
CyberShake_100010003988102.29
22.71
Inspiral_1000100032468.90
227.25
*When calculating the number of edges, average data size and average task runtime, the pseudo entry/exit VX-11e and the related edges are included.Full-size tableTable optionsView in workspaceDownload as CSV
Fig. 7. The structure of evaluated workflows. (a) Montage. (b) Epigenomics. (c) CyberShake. (d) Inspiral.Figure optionsDownload full-size imageDownload as PowerPoint slide
Besides anaphase real world workflows, we also test our algorithm on workflows generated by a DAG generator tool, which was also used in [49]. The tool defines the DAG shape based on four parameters: width, regularity, density, and jumps. In our experiment, for random DAG generation, the parameters are set as width = [0.2, 0.4, 0.8], regularity = [0.2, 0.4, 0.8], density = [0.2, 0.4, 0.8] and jumps = [1, 2, 3]. With these parameters, one DAG with 1500 jobs is created by choosing the value for each parameter randomly from the parameter data set.