Adaptive Execution of Scientific Workflow Applications on Clouds - Cloud Computing with e-Science Applications - page 76

FIGURE 4.1
Graphical representation of a simple scientific workflow.
were distributed, and thus data movement across wide distances might be
necessary. Even in this case, focus was still on execution time minimization.
Cloud computing adds a new dimension for workflow execution related
to the financial cost of using a virtually infinite amount of resources for
workflow execution. This means that the only limitations to the available
resources, and consequently the improvements in execution time, are the
available budget for workflow execution and the structure of the workflow itself,
which determines the maximum amount of tasks that can be executed in
parallel in the infrastructure. Clouds also brought other challenges for work-
flow management and execution. They are discussed in the next section.
4.3 Requirements for Adaptive Execution
of Workflows on Clouds
Although modern WfMSs already support clouds as the platform support-
ing the execution of workflow applications, many desirable features are still
absent in the WfMSs. This is because current WfMSs for clouds are derived
from projects in the area of grid computing. Therefore, many of their fea-
tures are optimized for grids and thus are unable to obtain the most key
aspects of clouds, such as rapid elasticity.
In this sense, clouds add extra complexity to WfMSs because the amount of
resources that WfMSs can provision for executing the workflow is virtually
infinite, as long as there is budget available to spend on the workflow execu-
tion process. Thus, different from existing algorithms and approaches that
operated with the goal of obtaining the most from the resources available for
the application, cloud-enabled WfMSs can assume that the main restriction
of the system is the budget rather than resources, and its goal is balancing
utilization, cost, and reduction of execution time [8].