FREE ELECTRONIC LIBRARY - Abstract, dissertation, book

Pages:     | 1 |   ...   | 16 | 17 || 19 | 20 |   ...   | 25 |

«Approximate Information Filtering in Structured Peer-to-Peer Networks Christian Zimmer Max-Planck Institute for Informatics Saarbrücken ...»

-- [ Page 18 ] --

[STSW02] commonly used in such a P2P search engine. A peer can select a random or query-specific document from this shared collection to store it in its local database. The shared collection is also realized by a Cloudscape database. The peer instances can run on one or more machines. The following sections present a usage scenario and explain the graphical user interface of the client in detail. Minerva Initialization

When starting the Minerva client, the user has to input the network details and database login as shown in Figure 6.4. On the left, the Local Port and the Nickname of the peer have to be specified. If the peer joins an existing network, the Remote IP address and the Remote Port number of a random peer in the network must be declared. On the right, the connection to a local database has to be stated including DB Service Name, Host Name, Port number, Username, and Password. Here, the database DB1 runs on the same machine realized by the server mentioned in the previous section.

The Create Ring button specifies that a new P2P network should be created whereas the Join Ring button is used to contact an existing network. The form also allows to select which widgets should be shown automatically after initialization. All three check boxes are preselected.

–  –  –

After initialization, the peer is connected to a P2P network. Figure 6.5 shows on the right the Network Properties and Collection Statistics. The Network Properties illustrate that the peer is connected, has a certain Pastry Node ID with local port, and owns the local collection with DB1 as DB Service Name. The Collection Statistics present some statistical information (e.g., the number of documents) concerning the local collection.

The Received Posts widget in the middle manages the metadata the peer is responsible for. The list shows all keys a peer stores metadata for, e.g., three peers in the network have published metadata for key music. The entry of Peer02 tells that this peer hosts 17 documents containing the key music. The Refresh button updates the shown list such that metadata received in the meantime from other network peers is updated. The Post all button starts the posting procedure of this peer and distributes the metadata of its local collection to the network. In addition, the Collection Statistics are recomputed to incorporate new documents that have been recently published. One-Time Query Execution

Figure 6.6 summarizes the one-time query execution.

The query input field contains the requested query modern music, and the unselected check box designates that this is a onetime query. When the user executes the query, the peer contacts the directory peers storing the metadata for key modern and key music to retrieve the key statistics. For this query, the peer itself hosts the key statistics such that it is fact that only Peer02 stores documents for both keys.

–  –  –

Figure 6.5: Updating and Disseminating Metadata.

Peer selection for this query is trivial and the one-time query is forwarded to this peer.

Nevertheless, Minerva allows to specify the peer selection strategy (e.g., CORI [SJCO02]) and the number of remote peers that receive the query. The Query Results widget shows the final result document list to the user. All three documents are hosted at Peer02. The URL allows to visit the document’s online version. If more than one peer contribute their local results for the requested query, and if the overall number of results is high, Minerva merges the collected result documents (e.g., using CORI-based merging algorithms) and displays only the top-ranked documents to the user. Continuous Query Execution

Subscribing with a continuous query using the Minerva prototype with extended MAPS functionality is shown in Figure 6.7. The user enters the continuous query in the text field mentioned before and selects the check box on the left to specify that this request is a continuous query. In the background, Minerva checks whether this continuous query is already active, i.e., it was requested by the same user in the past. If the query is active, the directory is used to retrieve updated statistical metadata about the query keys, and time series analysis is applied according to the MAPS approach presented in this thesis.

Prediction methods such as double exponential smoothing can not be applied to continuous queries requested for the first time. Thus, resource selection alone is used to select the most promising publishers in the future. Having selected the interesting publisher peers, the Minerva prototype system sends the continuous query to them and waits for new published documents. Periodically, continuous queries have to be updated to recognize publishing behavior. This can be realized automatically by the system or manually by the user.

–  –  –

Figure 6.8: Publishing Documents with Notifying Subscriber Peers. Document Publication Figure 6.8 illustrates the publication process for the showcase where a peer can publish new documents by adding them to its own local collection. Each peer in the network has a database connection to this server, and two different methods allow to publish new


• Random Insert selects a completely random document from the server collection and adds the selected one to the local database. A random document can not be published twice such that the server reminds of this publication.

• Matched Insert allows in this showcase to select documents that match active continuous queries a peer stores. In this case, the peer selects a random stored information demand and gets a matching document from the server to add it to its local store.

If there is no matching document available or no active continuous query stored, a random publication occurs.

On the lower right, Figure 6.8 shows the incoming requests. The first request was the one-time query for modern music and the second request was the same query as long-term demand. Having added new documents to the local collection, a peer updates its Collection Statistics and disseminates its refreshed metadata to the network by pushing again the Post all button. This procedure can be done periodically to decrease network traffic caused by update messages.

–  –  –

Receiving a notification for a new published document is shown in Figure 6.9. Here, the querying peer for modern music gets a notification message from Peer02. This message is included in the Notifications widget. Of course, the user can directly follow the URL link to access the online-version of the published document. In addition, the Running continuous queries widget lists all active subscriptions of the current peer. Again, continuous queries with expired lifetime are removed from the list. Resubmitting Continuous Queries

The last screenshot deals with the resubmission of already existing continuous queries.

Figure 6.10 shows that the query modern music is requested again.

In the meantime, several peers in the network have published new documents. The Received Posts widget lists all currently available metadata for the two requested keys. Now, two peers have published documents concerning modern, and five peers store data regarding music. Thus, peer selection for the whole query selects Peer01 and Peer02 also considering the time series observations since the query was last requested.

The Running continuous queries widget designates that this query was stored before, such that metadata of previous executions and selection processes are available to be used in the future.

–  –  –

Figure 6.10: Resubmitting a Continuous Query.

6.4 Other Prototypes This sections briefly introduces some existing prototypes for P2P retrieval and P2P filtering.

LibraRing [TIK05a] is the only other system that combines retrieval and filtering functionality in a P2P environment of digital libraries. In contrast to the prototype presented in this chapter, LibraRing focuses on exact searching and filtering functionality by disseminating documents or continuous queries in the network. Section 6.4.1 presents some P2P retrieval systems, whereas Section 6.4.2 discusses relevant P2P filtering systems. Overall, the prototype presented in this thesis is the only approach that provides approximate P2P searching and filtering functionality in a unifying framework.

6.4.1 P2P Retrieval Prototypes Galanx [WGD03, GWJD03] is a P2P search engine implemented using the Apache HTTP server and BerkeleyDB. Web site servers form the P2P layer of this architecture; pages are stored only where they originate from. Galanx directs user queries to relevant peers by consulting a local peer index that is maintained on each peer. In the experimental evaluation, the use of peer indices to direct searches is investigated. In contrast, the Minerva approach relies on peers to decide at what extent they want to crawl interesting fractions of the Web and build their own local indexes. [GWJD03] focuses on XML data repositories and postulates that, upon completion of the query, regardless of the number of results or how they are ranked and presented, the system guarantees that all the relevant data sources known at query submission time have been contacted. For this purpose, a distributed catalog service that maintains summaries of all peers is designed.

- 117 Chapter 6 Prototype Implementation

PlanetP [CAPMN03] is a publish-subscribe service for unstructured P2P communities, supporting content ranking search. PlanetP distinguishes local indexes and a global index to describe all peers and their shared information. The global index is replicated using a gossiping algorithm. PlanetP does not provide notification messages about new published data. Odissea [SMwW+ 03] (Open DIStributed Search Engine Architecture) assumes a two-layered search engine architecture with a global index structure distributed over the peers in the system. The system provides a highly distributed global indexing and query execution service that can be used for content residing inside or outside the P2P network. A single peer holds the complete, Web-scale, index for a given text key (i.e., keyword or word stem). Query execution uses a distributed version of Fagin’s threshold algorithm [Fag02].

The system appears to cause higher network traffic when posting document metadata into the network, and the presented query execution method seems limited to queries with at most two keywords. The paper actually advocates using a limited number of peers, in the spirit of a server farm.

The OverCite system [SCL+ 05] was proposed as a distributed alternative for the scientific literature digital library CiteSeer. This functionality was made possible by utilizing a DHT infrastructure to harness distributed resources (storage, computational power, etc.).

OverCite is able to support new features such as documents alerts. The work presented in [RV03] adopts an architecture very similar to Minerva, but seems incomplete since one cannot locate a running implementation. The presented results are based on simulations that also support the assumption that a Minerva-like architectures do in fact scale and are well within reasonable bandwidth limits. The system described in the paper provides keyword search functionality for a DHT-based file system or archival storage system, to map keyword queries to unique routing keys. It does so by mapping each keyword to a peer in the DHT that will store a list of documents containing that keyword.

The eSearch system presented in [TD04] is a P2P keyword search system based on a hybrid indexing structure in which each peer is responsible for certain keys. Given a document, eSearch selects a small number of important keys in the document and publishes the complete key list for the document to peers responsible for those top keys. This selective replication of key lists allows a multi-key query to be processed locally at the peers responsible for the query keys, but the document granularity indexes may interfere with the goal of unlimited scalability. The authors claim that eSearch is scalable and efficient, and obtains search results as good as state-of-the-art centralized systems.

Rumorama [EH05] is an approach based on the replication of peer data summaries via rumor spreading and multi-casting techniques in a structured overlay. Rumorama utilizes a hierarchical structure, and adopts a summary-based approach to support P2P-IR in the spirit of PlanetP. In a Rumorama network, each peer views the network as a small PlanetP network with connections to peers that see other small PlanetP networks. Each peer can select the size of the PlanetP network it wants to see according to its local processing power and bandwidth. Rumorama manages to process a query such that the summary of each peer is considered exactly once in a network without churn. The actual number of peers to be contacted for a query is a small fraction of the total number of peers in the network.

Alvis [LKP+ 05] is a prototype for scalable full-text P2P-IR using the notion of Highly Discriminative Keys (HDK ) for indexing, which claims to overcome the scalability problem of single-key retrieval in structured P2P networks. Alvis is a fully-functional retrieval engine built on top of P-Grid. It provides distributed indexing, retrieval, and a content-based ranking module. While the index size is even larger than the single key index, the authors bring forward that storage is available in P2P systems as opposed to network bandwidth.

ALVIS includes a component for HDK-based indexing and retrieval, and a distributed content-based ranking module.

Pages:     | 1 |   ...   | 16 | 17 || 19 | 20 |   ...   | 25 |

Similar works:

«KÖZJEGYZŐK KÖZLÖNYE KÖZJEGYZŐK KÖZLÖNYE A Magyar Országos Közjegyzői Kamara szakmai folyóirata 2014 / 6. szám 2014. november / december . oldal Juhász Gábor Az örökhagyó szabad rendelkezési jogának korlátjaként megjelenő kötelesrész a spanyol jogban (I. rész) . oldal Csillag Attila A közjegyző szerepe a francia örökösödési eljárásban . oldal Rádonyi Dénes A halál esetére szóló alapítványrendelésről . oldal Székely Erika...»

«SA GERMAN ASSOCIATION Inc. NEWSLETTER Das Band SÜD-AUSTRALISCHER ALLGEMEINER DEUTSCHER VEREIN INC. AprilMai 2013 Happy Easter to all our Easter Happy Members to all our Members SAADV (SA German Association Inc.) 223, Flinders Street, Adelaide, SA 5000, Australia Tel.: (08) 8223 2539 Email: office@thegermanclub.com.au Fax: (08) 8232 2082 Web: www.thegermanclub.com.au www.thegermanclub.com.au/about-us/special-interest-groups.php SAADV Das Band Imprint SAADV NEWSLETTER Das Band 2012 Bi-Monthly...»

«Brent Vine: Publications (January 2012) ARTICLES 1977. “On the Heptasyllabic Verses of the Rig-Veda”; Zeitschrift für vergleichende Sprachforschung (KZ) 91, 246-255. 1978a. “Nasalization in the Sara Languages”; Afrika und Übersee 61, 119-135. 1978b. “On the Metrics and Origin of Rig-Vedic ná ‘like, as’”; Indo-Iranian Journal 20, 171-193. 1981. “Remarks on African Shadow Vowels”, in Harvard Studies in Phonology 2 (ed. G. N. Clements), 383-427. 1984. “African ‘Shadow...»

«www.thecustomerexperience.es contents experiences 6 foreword 9 intro 12 Customer experience: a framework for the marketing of the future Elena Alfaro Partner EMO Insights brand and communication 20 Customer experience from the perspective of the brand and communication Javier Velilla Managing Partner COMUNIZA management systems 29 The role of IT systems in managing customer experience Hugo Brunetta CEO Nexting measurement 38 How to measure customer experience Carlos Molina Innovation Vice...»

«AASB Exposure Draft ED 182 June 2009 Prepayments of a Minimum Funding Requirement [AASB Interpretation 14] Comments to AASB by 13 July 2009 Commenting on this AASB Exposure Draft Constituents are strongly encouraged to respond to the AASB and the IASB. The AASB is seeking comment by 13 July 2009. Comments should be addressed to: The Chairman Respondents to the IASB are asked to send Australian Accounting Standards Board their comments electronically through the PO Box 204 ‘Open to Comment’...»

«Jobst C. Knigge Feltrinelli – Sein Weg in den Terrorismus Humboldt Universität (open access) Berlin 2010 INHALTSVERZEICHNIS EINLEITUNG HERKUNFT Eine der reichsten Familien Italiens Die Eltern Der Millionenerbe SUCHE NACH DEM EIGENEN WEG Politische Anfänge Der eigene Verlag Das Feltrinelli-Institut Enttäuschung über die KPI Der Fall „Doktor Schiwago” „Der Leopard” – Nostalgie eines Landadligen Erfolgreiche Jahre des Verlages Vier Ehefrauen KUBA UND GUERILLA Kuba Bolivien Vorbild...»

«Unknown Book 7594416 If spending the action it can have your times are. Very, of the staffing user with interest and good achievement beginning for a secured customer, you will keep credit Unknown Book 7594416 as getting garnishments greet in a high-traffic you are also drastically. Consulting offered them started fixing to consider the year! Getting at frontiers can already but then deal does give-up rent of care. Us need the to-do, giving ingredients before bucking the cost, and you have the...»

«DGPF Tagungsband 24 / 2015 Wasserlinienextraktion aus optischen Nahbereichsaufnahmen mittels Texturmessverfahren MELANIE KRÖHNERT1, ROBERT KOSCHITZKI1 & HANS-GERD MAAS1 Zusammenfassung: In diesem Beitrag wird ein Verfahren zur Detektion von Wasserlinien offener Gewässer mithilfe optischer Nahbereichsaufnahmen auf Basis von Texturund Spektralinformationen vorgestellt und evaluiert. In zwei Studiengebieten unterschiedlicher Topographie wurden mit handelsüblichen Amateurkameras Studien im...»

«Gegenständliche Modellierung virtueller Informationswelten Eva Hornecker und Kai Schäfer Forschungszentrum Arbeit und Technik (artec), Universität Bremen Veröffentlicht in: Software Ergonomie ´99, U. Arend, E. Eberleh, K. Pitschke (Hrsg.), Teubner, Stuttgart, 1999 Zusammenfassung In diesem Beitrag stellen wir ein Konzept greifbarer, gegenständlicher Benutzungsschnittstellen vor. Es werden Übergänge zwischen dem Modellieren im Realen und dem Erstellen virtueller Modelle geschaffen....»

«A Móra Ferenc Múzeum Evkönyve 1972—73/1 AZ APÁTFALVI NÉP TÁPLÁLKOZÁSA II. SZIGETI GYÖRGY (Szeged, Cs. M. Művelődésügyi Módszertani Tanács) DISZNÓÖLÉS APÁTFALVÁN Apátfalván azt tartották, hogy december első napjaitól bármikor el lehetett kezdeni a disznóvágást. A legtöbb disznót mégis közvetlenül karácsony előtt, vala­ mint karácsony és újév között vágták, amikor már hidegebb volt az idő és az ünne­ pek alatt a család is jobban együtt...»

«EVLİYÂ ÇELEBİ Studies and Essays Commemorating the 400th Anniversary of his Birth EDITORS Nuran Tezcan · Semih Tezcan Robert Dankoff REPUBLIC OF TURKEY MINISTRY OF CULTURE AND TOURISM PUBLICATIONS © THE BANKS ASSOCIATION OF TURKEY © REPUBLIC OF TURKEY MINISTRY OF CULTURE AND TOURISM GENERAL DIRECTORATE OF LIBRARIES AND PUBLICATIONS Republic of Turkey Ministry of Culture and Tourism 3358 The Banks Association of Turkey Publications General Directorate of Libraries and Publications 290,...»

«Ontologie-basierte Hypertextsorten-Klassifikation Georg Rehm 1 Einleitung Der breite Einsatz computerlinguistischer Verfahren in Projekten zur Bewältigung der häufig zitierten Informationsflut beschränkt sich bislang meist auf Vorverarbeitungsprozesse (z.B. Wortstammreduktion oder Wortartenannotation zur Verbesserung von Information Retrieval-Algorithmen) oder klassische Anwendungen wie das automatische Textzusammenfassen oder die maschinelle Klassifikation eines Webdokuments in ein...»

<<  HOME   |    CONTACTS
2016 www.abstract.xlibx.info - Free e-library - Abstract, dissertation, book

Materials of this site are available for review, all rights belong to their respective owners.
If you do not agree with the fact that your material is placed on this site, please, email us, we will within 1-2 business days delete him.