Information Technology Reference
In-Depth Information
by W3C (2000, 2003) for exposing the low level
functionality on a network. The recent advance-
ment to WS-RF (http://www.globus.org/wsrf/)
provides decoupled mechanisms for representing
stateful
resource capabilities
through stateless
Web services interfaces, which allows a sys-
tem to manage lifetime without using a tightly
coupled approach like previous distributed object
systems, such as Corba (http://www.corba.org/)
or Jini (http://www.jini.org/). Grid computing is
built on standardized technologies and extended
through other standardization efforts often initi-
ated through the various research and working
groups hosted within the Open Grid Forum (OGF)
conference (http://www.ogf.org/).
Peer-to-peer (P2P) technologies on the other
hand, have grown through grass roots Internet
culture, initiated by specific applications such as
file sharing systems like Napster (http://www.nap-
ster.com/) or Gnutella (http://gnutella.wego.com/)
and CPU sharing systems like SETI (Search for
Extra Terrestrial Intelligence (http://setiathome.
ssl.berkeley.edu/)). At its essence, P2P is about
connecting people so that users can share informa-
tion or participate in projects by being offered a
various range of incentives to do so. P2P therefore
is much more focused on solving scalability and
robustness issues when faced with huge numbers
of connected participants, which are extremely
unreliable in that they can disconnect frequently
and are hosted behind application-unfriendly
mechanisms, such as Network Address Transla-
tion (NAT) systems and firewalls. With such a high
number of highly transient participants, writing
applications for such a networking environment
requires various design issues to be tackled in
order to cope with such a dynamic environment,
for example, scalability, reliability, interoper-
ability, and so on.
The growth of these two largely independent
areas has been exponential with enormous mo-
mentum and diversity. However, there is also an
obvious convergence as production Grids move
forward into deployment over significantly more
participants. The scalability techniques offered
through P2P algorithms not only balance the
load through encouraging the use of
servents
,
where each peer is both a client and a server, but
also address robustness through the use of more
decentralized network overlays that can adapt to
random failures across the network.
In this chapter we describe the DART (Dis-
tributed Audio Retrieval using Triana) project
that is being researched jointly between Cardiff
University and the Laboratory for Creative arts
and Technologies (LCAT) in Louisiana State
University that capitalises on these developments
to provide a decentralized overlay for the process-
ing of audio information for application in Music
Information Retrieval (MIR). The design of the
system is based on a BOINC-like (http://boinc.
berkeley.edu/) paradigm; similar to the way
SETI@home works. Under these scenarios, us-
ers provide CPU cycles to projects to help in the
analysis of data for searching for both scientific
and non-scientific events. Within DART we use
the same type of philosophy to enable users to
provide metadata to the network that is created
through the analysis of the audio files located on
their individual hard drives. The local processing
on the network participants is achieved though the
use of a popular Grid and P2P enabled workflow
environment, called Triana, which allows the
specification of data flows, to provide pipelines
for processing the data in specific ways. Triana
enables extremely rapid prototyping and code
reusability of analysis algorithms exposed by a
graphical interface, where units (representing
audio analysis tools) can be dragged, dropped
and connected together in order to create the
algorithms. Such algorithms can be deployed
that is, uploaded to the peers in the network,
through the use of decentralized P2P structures
that can enable caching and file replication for
speed and efficiency. Although, in an early stage
of development, DART provides insight into the
aggregation of P2P, Grid and MIR technologies to
Search WWH ::
Custom Search