Information Technology Reference
In-Depth Information
BOINC in that it exploits the presence of multiple
data centres in order to replicate input data files
across the Grid network.
will always be evolving. Initially, it is easier to
concentrate on statistical methods for recom-
mending music to the users of the DART system
(searching for correlations between artists found
on the users machines), however this will soon
evolve, and become more involved as the Triana
workflow is refined and more MIR algorithms are
incorporated into the package sent to the worker
nodes. The algorithms used for the recommenda-
tion of music files to the DART users will be made
available to the MIR community for refinement,
suggestions, and also to allow for the advancement
of this field of research as MIR-algorithm special-
ists provide input ideas and offer improvements
to refine and maximise the benefit of the use of
the DART system.
peer overview
This section gives a brief overview of what role
each node on the network plays, and the jobs as-
sociated with that role:
DART Manager
•
Creates new task graphs and Triana units
•
Advertises new DART Packages
User
dart flexibility
•
Queries Results
•
Downloads/Receives Results
The chapter so far has focussed on the creation
of a distributed P2P MIR system to facilitate the
recommendation of music to users intelligently,
and the framework created in order to achieve
this. However the DART infrastructure can be
used for a variety of applications, musical, and
otherwise. One application scenario is a DART
system installed on a local/closed network; scan-
ning sound effect audio files on separate systems
in networked commercial studio facilities, to
search for sound effect files (or for example,
drum sounds) that match a specific criteria. For
example, if a DART user requires a 'snare' drum
sample, usual search mechanisms cannot search
for suitable sounds unless the audio files filename
contains the word “snare” or other suitable identi-
fier—the search mechanism would do no more
than string matching.
In contrast to this, given the appropriate algo-
rithm created in Triana by the DART manager,
then propagated onto the network, the DART
system would allow the user to search all the au-
dio content in the distributed sound library, and
suggest files with suitable characteristics which
match the specific search criteria set by the user.
Worker Nodes
•
Queries New Packages
•
Downloads Package
•
Works/processes Taskgraph
•
Advertises Results
Package Repositories
•
Query and Download New Packages (Stored
locally)
•
Advertise Packages
Super Peers
•
Caches Advertisements
•
Does simple analysis of results received
from workers
Workflow Design
The workflow created by the DART manager and
propagated onto the other peers on the network,
Search WWH ::
Custom Search