Information Technology Reference
In-Depth Information
PsC Model
In both the previously mentioned models,
the role of the coordinator can significantly ease
the process of retrieval due to its supervising
character in terms of route to peers discovery
and knowledge of the available content on other
peers (PsCM). Though, centralized P2P networks
are subject to the same drawbacks for which the
traditional server-client model was originally not
used (network failures due to central peer failure,
impaired scalability, joining/leaving of peers not
easily handled and possible undesirable dominion
of coordinator controllers). In other words the
coordinator is easily overloaded and becomes
the bottleneck of the whole system, leading to
poor stability. As for these two models (PsCM
& PsC+M) Wang et al. (2002) conclude that the
load for a given query produced by PsCM is less
than that of PsC+M. Though comparing time re-
quirements for the completion of the same query
PsC+M is faster than PsCM.
In the peers-coordinator model (PsCM) peers
contain a music set as well as sharing a common
feature extraction method. A coordinator main-
tains a data structure to store all music features in
the P2P system and a feature matching method to
compute the distance between two music features.
Queries are content requests proposed by a user.
In this model, any query PC in the system con-
nects with the coordinator, while peers connect
with each other via the coordinator, though they
directly transfer music data from one to another.
During the query process the feature of the music
request is extracted from the music request by an
extraction method and sent to the coordinator.
The coordinator receives the feature, compares
it with all music features uploaded during peers
registration and sends the result to the request PC.
Results contain the locations of the peers which
store the music similar to the requested.
decentralized structured
PsC+ Model
In the system proposed by Tzanetakis et al. (2004)
each node in the network stores music files for
sharing as well as information about the location
of other music files in the network. Shared files
in the system are described by a Music File De-
scription (MFD) which is essentially a set of AV
pairs (artist = “Jennifer Warnes”, album = “Dirty
Dancing”, song = “The Time Of My Life”, ...,
specCentroid = 0.65, mfcc2 = 0.85, ...). A Music
Feature Extraction Engine (MFEE) calculates
the features from the audio files, while others are
manually appointed. Two operations are supported
by the system, the registration of a music file based
on its associated MFD and the search where the
user query is converted into an appropriate MFD
that is then used to locate nodes containing files
matching the search criteria. The MFD of either
registration or search query is passed to the content
discovery system (CDS), which runs on top of a
A slightly differentiated implementation of a
centralized system, PsC+M, assigns the feature
matching process on each peer. In the peers-
coordinator+ model (PsC+M) each peer has ad-
ditionally to the PsC model a commonly shared
feature matching method. The coordinator con-
sists of a data structure which stores the network
identifiers of all peers. The main steps of a query
process are the following: Each peer shares some
music stored on the local hard disk and registers
its location at the coordinator. A query by any
peer sends the coordinator the extracted feature
of the music request. The coordinator forwards
the feature of the query to all registered peers,
each peer compares it with the features of the
local shared music and sends the local result to
the coordinator. The coordinator in turn passes
on the results to the querier.
Search WWH ::




Custom Search