Information Technology Reference
In-Depth Information
structures may be examined with an integrated 3D molecule viewer based on
ChemDoodle (2014). Numerical data, in particular plots and graphs, can be
inspected via a dygraphs implementation (Dygraphs 2014). The latter two are based
on JavaScript and do not require any further installation of Java plugins.
11.2.3 Distributed Data Management
The MoSGrid science gateway infrastructure builds on distributed data manage-
ment for storing the simulation data of its users. Thus, simulation data is kept
independently of the location of the calculation. The infrastructure is designed to be
modular and different
file systems can be supported.
XtreemFS is a distributed object-based
file system for grid and cloud environ-
ments. In MoSGrid, XtreemFS has been integrated to enable redundant and dis-
tributed storage of molecular simulation data. The data objects, chunks of a given
file, are managed by object storage devices (OSDs), and information such as the
directory tree, sizes, timestamps, and ownership is stored by the metadata and
replica catalog (MRC). The
file system can be accessed via a Filesystem in User-
space (FUSE) module (FUSE 2014), which offers a POSIX
file system interface or
a Java API. Both X.509 certificates and SAML trust delegation assertions are
supported for authentication. Within MoSGrid, Java API and SAML support was
developed.
In the science gateway the
file system is accessed via the Java API and a SAML
assertion. Users are able to choose input
files during the workflow parameterization
and to access
files during or after the execution of a workflow. Finally, the MoS-
Grid data repository can be browsed to access past and present simulations. Files
stored in XtreemFS can be addressed via the XtreemFS URL schema, xtreemfs://.
When UNICORE registers this schema in an import or export
field in a job defi-
-
nition the
file is accessed via XtreemFS. This integration is done by running the
XtreemFS FUSE client on a cluster login node, and when UNICORE accesses
the mounted directory in the name of a user, it can import or export
files to/from the
UNICORE job working directory. Due to the modular approach of MoSGrid,
additional distributed
file systems can easily be integrated.
11.3 Metadata Management
MSML serves as the central MoSGrid data format and it signi
cantly eases the
handling of chemical
file formats and results. Developed parsers and perform
conversions from chemical input formats to MSML, output formats to MSML, and
MSML back to chemical input formats. Additionally, MSML in conjunction with
UNICORE is used to enable the user to easily search for and
find any MoSGrid
related data.
Search WWH ::
Custom Search