A Data Stream View of Scientific Workflow - Grid Computing: Infrastructure, Service, and Applications - page 223

9.1.4.2 How to Build up a WFMS with DBMS Views

Based on the introduction of the transactional and objective views for
workl ows, below is a procedure for implementing a WFMS through the
objective database view according to [6].

• A workl ow is described as an object-oriented database schema

• An instance of the workl ow schema is created during execution

• Invocation of workl ow processes is captured and activated by

rules from a constraints form
External applications implementing workl ow processes are •

scheduled through updates to system catalogs
Status information and other kinds of information about running •

or i nished workl ow processes are obtained by database queries
Information on the workl ow data is obtained by database queries •

9.1.5 Data Stream and Scientific Workflow

There exist overlapping research issues not only between the scientii c
workl ow management systems and DBMS, but also between the scientii c
workl ow management system and the data stream management system
(DSMS) [14,15]. For example, the workl ow execution and scheduling
are somehow similar to the generation of the data stream query plan.
Furthermore, the control l ow and data l ow [16] themselves can be treated
as data streams, whose tasks focus on raw data prereduction, analysis,
calculation, selection, and i ltering. Even within the range of the pattern
and representation, the processing techniques from DSMS can enrich the
content in scientii c workl ow. Finally, the real-world impacts for scientii c
workl ow applications breathtakingly resemble those for data streams
such as meteorology, astronomy, and monitoring sensor network. Sponta-
neously, it comes into the picture that a DSMS view can be taken into the
scientii c workl ow area.

9.1.5.1 Data Stream Brevity

For the strangers to DSMS with skeptical eyes, the i rst question is what is
a data stream and what is a DSMS? Generally speaking, a data stream can
be described as multiple transient relations or structures, each of which
consists of ini nite tuple sequence or dei ned data items. The data update
method of data streams is append only. With respect to query processing,
queries on data streams are usually continuous with approximate answers.
In addition, query evaluation takes the form of one pass, and the query
plan is adaptive. A DSMS is designed to manage or i lter a large body and