Information Technology Reference
In-Depth Information
process but also from scientii c and engineering computing; for example,
bioinformatics, high energy physics, and astronomy. This requires the
grid workl ow systems to support new types of application. Furthermore,
multiple grid middleware needs to be integrated into the workl ow sys-
tem, which is supposed to run on heterogeneous grid infrastructures. The
research challenges and issues of grid workl ow that have emerged
in comparison with traditional business workl ow environment are dis-
cussed as follows [59]:
•
Workl ow applications running on computational grids usually
manipulate large amounts of data, which are of various types and
from heterogeneous data sources. The grid workl ow should
support the efi cient transfer of large amounts of data and resource
discovery for heterogeneous data sources.
Lifecycles of grid workl ow applications may have a wide range;
•
for example, from one minute to dozens of days. This brings
challenges for fault tolerance, application monitoring, and steer-
ing, for example.
The grid infrastructure and environment, including APIs, inter-
•
faces and protocols, are different from traditional workl ow envi-
ronment. A number of legacy codes and applications are not
suitable for running on computational grids. Usually, these legacy
codes need to be wrapped as Web services before integrating into
the workl ow.
Some applications [16] running on computational grids need a real-
•
time support environment; for example, real-time resource alloca-
tion and reservation. There are few real-time requirements for
traditional business workl ows. The workl ow management system
should be designed to meet the requirements of real-time support.
The computational grid is a highly dynamic environment; for
•
example, CPU load, network bandwidth, and resource availabil-
ity. The dynamic feature brings difi culties for resource allocation
and application steering for workl ow system applications.
The computational grid covers multiple administrative domains
•
and lacks central control and complete information, which is not
needed in traditional workl ow enactment.
11.3.3
Start-of-the-Art: Service Component Architecture
The service-oriented architecture (SOA) approach promotes building
systems or workl ows from distinct and autonomous Web services. By
using SOA to create the service-oriented workl ow, the autonomous busi-
ness logic or legacy scientii c code can be presented as a service; hence, the
Search WWH ::
Custom Search