Database Reference
In-Depth Information
designed, modified, or enhanced. As a result the users may retrace their steps
in the design process, choose various “flavors” of the same workflow and try
and retry different designs. A challenge could be not only to capture the
how
but also the
why
of the design decisions made by the users.
Unlike metadata, much process documentation is relatively easy to produce
automatically, especially in the context of workflows, since the workflow sys-
tem is in charge of setting up the environment for the computation, managing
the data, and invoking the analysis steps. Thus, the workflow system is able to
capture the information about where the execution took place, what were the
parameters and environment variables used by a software component, which
data files were used, and so forth. Some of that information may be somewhat
obscured, for example, when a configuration file is used instead of placing the
parameters on the command line. However, the workflow system can also au-
tomatically save information about which configuration file was used. It is
also interesting to note that the capabilities of workflow management and
provenance management systems are complementary (process execution vs.
process documentation), and thus it is possible to integrate workflow manage-
ment systems with provenance management systems that have been developed
independently of each other.
22
12.4 Survey of Existing Approaches
In this section we describe some of the existing approaches for managing
metadata and provenance. We start by giving an example of the various types
of attributes that are part of the metadata of scientific data.
12.4.1 Metadata Schema and Metadata Attributes
Metadata attributes that are elements of a metadata schema can encompass a
variety of information. Some metadata is application independent, such as the
creation time, and author, as described in Dublin Core
23
; and other metadata
is application dependent and may include attributes such as the duration of an
experiment, temperature of the device, and others. Many applications have ex-
panded the Dublin Core schema to include application-dependent attributes.
24
Based on experiences with a number of scientific applications, we de-
scribed nine general types of metadata attributes.
25
,
26
These descriptions
were used in metadata systems developed as part of the Metadata Cata-
log Service (MCS),
25
the Laser Interferometer Gravitational-Wave Observa-
tory (LIGO) project,
27
the Linked Environments for Atmospheric Discovery
(LEAD) project,
28
and others. Below we describe some of the metadata at-
tribute categories.
