Gridbus Workflow Enactment Engine - Grid Computing: Infrastructure, Service, and Applications - page 139

between the parent tasks' end time and the child tasks' start time. This
gap can be attributed to the overhead of running WFEE, including time
involved in processing event notii cations, resource discovery, and remote
resource submission. However, compared with the running time of tasks,
this gap is insignii cant and less than 2%.
5.11 Related Work
The workl ow engine presented in this chapter is an independent work-
l ow execution system and takes advantage of various middleware services
such as security, grid resource access, i le movement, and replica manage-
ment services provided by the Globus middleware [6,20], and multiple
middleware dispatchers provided by the Gridbus broker.
Much effort toward grid workl ow management has been made.
DAGMan [21] was developed to schedule jobs to the Condor system in an
order represented by a DAG and to process them. With the integration of
Chimera [22], Pegasus [23] map, and execute complex workl ows based
on full-ahead planning. In Pegasus, a workl ow can be generated from a
metadata description of the desired data product using AI-based planning
technologies. The Taverna project [24] has developed a tool for the compo-
sition and enactment of bioinformatics workl ow for the life science com-
munity. The tool provides a graphical user interface for the composition of
workl ows. Other workl ow projects in the grid context include UNICORE
[25], ICENI [26], Karajan [27], Triana [28], and ASKLON [29].
Compared with the work listed above, the workl ow engine provides a
decentralized scheduling system by using the tuple spaces model, which
facilitates deployment of different scheduling strategies to each task. It also
enables resources to be discovered and negotiated at runtime.
A number of workl ow languages [7,30,31] have been developed and
most of them focus on the composition of Web services. However Web
services are not the standard middleware used by the majority of today's
scientii c domains [32]. The workl ow language proposed in this chapter
is middleware independent and also supports parameterization [9], which
is important to scientii c applications.
5.12 Summary
In this chapter, a workl ow enactment engine is introduced to facilitate
composition and execution of workl ows in a user-friendly manner. The
engine supports different grid middleware as well as runtime service