Ganzseitiges Foto

Transcription

Ganzseitiges Foto
INHALT
Informatik
Spektrum
Band 35 | Heft 2 | April 2012
Organ der Gesellschaft für Informatik e.V. und mit ihr assoziierter Organisationen
EDITORIAL
C. Müller-Schloer, H. Schmeck, T. Ungerer
71
Organic Computing
74
MITTEILUNGEN
Mitteilungen der Gesellschaft für Informatik 214. Folge
HAUPTBEITRÄGE
I. Scholtes, C. J. Tessone
75
Organic Design of Massively Distributed Systems:
A Complex Networks Perspective
87
Cautionary Tales: Lessons Learned About
Unanticipated Behaviors in OC Systems
93
Selbstorganisierende Smart-Kamera-Systeme
P. R. Nelson, M. H. Nguyen, C.-C. J. Yang
J. Hähner, U. Jänen, C. Grenz, M. Hoffmann
A. Kamper, H. Schmeck
102
Adaptives verteiltes Lastmanagement in Bilanzkreisen
M. K. Müller, M. Tremer, C. Bodenstein, R. P. Würtz
112
Lernen situationsunabhängiger Personenerkennung
159
Video inpainting
J.-P. Steghöfer, W. Reif
119
Die Guten, die Bösen und die Vertrauenswürdigen –
Vertrauen im Organic Computing
AKTUELLES SCHLAGWORT
M. Atzmueller
132
Mining Social Media
DUECK- β-INSIDE
G. Dueck
136
Untrolle in Meetings und im Leben
HISTORISCHE NOTIZEN
F. L. Bauer
141
Die ALGOL-Verschwörung
FORUM
150
160
Die Vorstandskolumne/Dagstuhl Manifesto/Whistleblowing – Recht oder Pflicht?/
Gewissensbits – Wie würden Sie urteilen?/Heinz Schwärtzel-Dissertationspreis/
Zum Turing-Jahr/Zum Titgelbild
MITTEILUNGEN
Mitteilungen der Gesellschaft für Informatik 214. Folge
Aus Vorstand und Präsidium/Unsere Fördermitglieder/Aus den GI-Gliederungen/
Tagungsberichte/LNI-Neuerscheinungen
166
GI-Veranstaltungskalender
168
SI-Mitteilungen und SI-Veranstaltungskalender
123
EDITORIAL
C. Müller-Schloer
H. Schmeck
}
T. Ungerer
Organic Computing
Organic Computing (OC) hat sich das Ziel gesetzt, Lösungen für zukünftige hochkomplexe IT-Systeme und deren Beherrschung zu entwickeln. Komplexität entsteht durch die sehr große Anzahl von Komponenten und Subsystemen
solcher Systeme, durch deren zunehmende Vernetzung und damit wechselseitige Abhängigkeit, durch die Offenheit
künftiger Systeme sowie durch die prinzipielle Unmöglichkeit einer exakten Vorhersage aller künftigen Ziele und Störeinflüsse. Veränderungen sind nicht die Ausnahme, sie sind die Regel. Und diese permanente Veränderung ist keine
Eigenschaft, welche wir diesen Systemen ohne Not verleihen. Veränderung ist unvermeidbar. Als Kernthese des OC
wurde daher formuliert:
Es ist nicht die Frage, ob adaptive und selbstorganisierende Systeme entstehen, sondern wie wir sie beherrschen.
Auch wenn Organic Computing noch eine junge Forschungsdisziplin ist, so kann es doch bereits auf eine ansehnliche
Geschichte zurückblicken. Nach dem in mehreren Workshops in den Jahren 2002 und 2003 erarbeiten Positionspapier
der GI und ITG [1] hat die OC-Gemeinschaft vielfältige Aktivitäten entwickelt. Das ursprünglich noch sehr visionäre
Konzept, welches als ,,Komplexitätsbeherrschung durch Selbstorganisation“ umschrieben werden kann, wurde in
einer Vielzahl von Forschungsprojekten konkretisiert und mit Leben gefüllt. Die zentrale Rolle spielte dabei das DFGSchwerpunktprogramm 1183 Organic Computing, in dem von 2005 bis 2011 in drei Zweijahresphasen jeweils etwa
20 Projekte gefördert wurden. Das Schwerpunktprogramm wurde im September 2011 abgeschlossen. Die Ergebnisse
sind im ,,OC Compendium“ auf 627 Seiten dokumentiert [2].
Das Schwerpunktprogramm Organic Computing verfolgte das Ziel, Konzepte zur Beherrschung zukünftiger hochkomplexer Informations- und Kommunikationssysteme zu entwickeln. Hierzu wurden zentrale Begriffe wie Adaptivität,
Autonomie, Robustheit, Flexibilität, Selbstorganisation und Emergenz formal untersucht, um das theoretische Problemverständnis zu vertiefen. Zudem lag ein besonderer Fokus auf Architekturen und Methoden für den Systementwurf, sowie auf deren Umsetzung in konkreten Anwendungsszenarien. Unter Architekturgesichtspunkten spielt
die im Schwerpunktprogramm entwickelte generische Observer/Controller-Architektur eine zentrale Rolle. Sie unterstützt die gesteuerte Selbstorganisation technischer Systeme und kommt inner- und außerhalb des Schwerpunkts
in verschiedenen Anwendungen erfolgreich zum Einsatz. Weitere im Schwerpunkt entwickelte Architekturansätze
umfassen organische Middleware-Systeme, verifizierbare Entwurfsmuster und fehlertolerante Robotikarchitekturen.
Für viele der entwickelten Architekturen sind maschinelle Lernverfahren bedeutsam, da sie eine autonome Anpassung der organischen Systeme an sich dynamisch verändernde Anforderungen der Einsatzumgebung unterstützen.
Arbeiten in diesem Bereich behandeln den Wissenstransfer zwischen Agenten ebenso wie Konzepte zum Lernen in
performanz- und sicherheitskritischen Umgebungen. Darüber hinaus wurden etablierte Lernverfahren wie Learning
Classifier Systeme erfolgreich erweitert und verbessert. Die entwickelten Methodiken wurden u. a. in dynamisch rekonfigurierbaren, fehlertoleranten Hardware-Plattformen, in adaptiven Verkehrssystemen und in Produktionsszenarien
eingesetzt.
DOI 10.1007/s00287-012-0599-2
Informatik_Spektrum_35_2_2012
71
{ EDITORIAL
Eine Vielzahl von OC-Aktivitäten auch außerhalb des Schwerpunktprogramms dokumentiert die Lebendigkeit
des Arbeitsgebiets. Es existieren mehrere Nachfolgeprojekte mit DFG- und BMBF-Förderung, in denen OC-Methoden
eingesetzt und weiterentwickelt werden. OC konnte in den sechs Jahren des Schwerpunktprogramms auch im internationalen und im nicht-universitären Bereich etabliert werden. So wurde das Themenfeld Organic Computing
fester Bestandteil wichtiger internationaler Konferenzen wie der International Conference on Architecture of Computing Systems (ARCS), der International Conference on Autonomic and Trusted Computing (ATC), der International
Conference on Biologically Inspired Cooperative Computing (BICC) und der International Conference on Autonomic
Computing (ICAC). Darüber hinaus wurde das Forschungsfeld durch eine Vielzahl von Workshops, Special Sessions und
Tutorials im Rahmen weiterer Tagungen repräsentiert. Die internationale Vernetzung führender Wissenschaftler wurde
durch die Organisation von Dagstuhl-Seminaren in den Jahren 2006 (Organic Computing – Controlled Emergence),
2008 (Organic Computing – Controlled Self-organization) und 2011 (Organic Computing – Design of Self-Organizing
Systems) kontinuierlich gefördert. Zudem wurde eine IEEE Computational Intelligence Society (IEEE CIS) Task Force
ins Leben gerufen, die auf eine verbesserte Sichtbarkeit des Themenfelds Organic Computing hinarbeitet. Weiterhin
befindet sich in den Fachbereichen Technische Informatik der GI und der ITG die Gründung einer Fachgruppe Organic
Computing in Vorbereitung.
Besonders erfreulich ist, dass mittlerweile bereits drei Organic-Computing-Professuren in Deutschland existieren.
OC befindet sich nach Abschluss des Schwerpunktprogramms in einer entscheidenden Phase. Wir sehen zwei
wesentliche einander ergänzende Ausrichtungen der künftigen Arbeit:
1. OC soll verstärkt in konkreten Anwendungen zum Einsatz kommen, und
2. es besteht auf mehreren Spezialgebieten des OC noch großer Forschungsbedarf.
Besonders interessante Anwendungsgebiete, für die z. T. auch schon prototypische Lösungen existieren, sind Verkehrstechnik, Kommunikationsnetze, Smart-Grid-Techniken, Fabrikautomatisierung sowie Robotik. Forschungsschwerpunkte
des OC werden gesehen im Bereich der Selbst-Optimierung (zur Laufzeit), der kooperativen und institutionellen Strukturen (Social OC) und der kognitiven Architekturen. Weiterhin fehlen noch weitgehend systematische Ansätze zum
Entwurfsprozess von OC-Systemen, welcher sich grundlegend von dem herkömmlicher Systeme unterscheidet, sowie
zur Einbeziehung des menschlichen Nutzers in das OC-System.
Wir haben uns für den Themenschwerpunkt in diesem Heft des Informatik-Spektrums auf Arbeiten konzentriert,
welche außerhalb des Schwerpunktprogramms laufen und somit wegweisend für die weitere Entwicklung des Organic
Computing sind. Dabei kann selbstverständlich keine vollständige Abdeckung aller interessanten aktuellen OC-Projekte
erreicht werden. Die Artikel haben alle einen eindeutigen Anwendungsbezug. Sie zeigen, dass OC-Techniken praktisch
einsetzbar sind. Die Anwendungen liegen in den Bereichen Networks, Robotik, Sicherheit, Smart Grid, Desktop Grid
Computing und Gesichtserkennung. Darüber hinaus werden auch grundsätzliche Fragen des OC angesprochen, welche
für die weitere Forschung richtungweisend sein können: Dies sind (1) der Einsatz von Methoden der statistischen
Mechanik (I. Scholtes), (2) die Einführung explorativen Verhaltens von OC-Systemen (P. Nelson) und (3) die Nutzung
sozialer Prinzipien wir Vertrauen und Reputation für die Selbstorganisation technischer Systeme (J.-P. Steghöfer). – Im
Einzelnen:
– Ingo Scholtes et al. von der ETH Zürich schlagen in ihrem Beitrag ,,Organic Design of Massively Distributed Systems:
A Complex Networks Perspective“ die Nutzung der Random Graph Theory nicht nur zur Analyse bestehender Netzwerke sondern auch als konstruktive Maßnahme vor, um zur Laufzeit Netzwerkstrukturen (Overlays) mit definierten
Eigenschaften wie Konnektivität, Fehlertoleranz, Durchmesser etc. zu generieren.
– Phyllis Nelson et al. von der California State Polytecnic University in Pomona, CA, berichten im Beitrag ,,Cautionary
Tales: Lessons Learned About Unanticipated Behaviors in OC Systems“ über unerwartete Effekte bei der Implementierung von OC-Mechanismen in realen Robotik-Anwendungen und schlagen den Einbau eines ,,opportunistischen
Explorationsverhaltens“ vor, welches es den Robotern – in Analogie zu Lebewesen – ermöglicht, ihre Welt- und
Selbstmodelle aktiv zu verbessern.
72
Informatik_Spektrum_35_2_2012
– Jörg Hähner et al. (Leibniz Universität Hannover) zeigen in ihrem Beitrag ,,Selbstorganisierende Smart-KameraSysteme“ auf, wie OC-Techniken genutzt werden können, um ohne eine zentrale Instanz eine fortlaufende
Optimierung der räumlichen Partitionierung auch unter Störungen wie dem Ausfall von Kameras zu erreichen.
– Andreas Kamper et al. (Cirquent GmbH bzw. KIT) diskutieren im Beitrag ,,Adaptives verteiltes Lernen in Bilanzkreisen“ die Anwendung mehrstufiger Observer/Controller-Strukturen für den selbstorganisierenden Lastausgleich in
Smart-Grid-Systemen und zeigen, dass sich so Erzeuger/Verbraucher-Ungleichgewichte deutlich reduzieren lassen.
– Rolf Würtz et al. von der Ruhr Universität Bochum berichten in ihrem Beitrag ,,Lernen situationsunabhängiger
Personenerkennung“ über den Einsatz von OC-inspirierten Lerntechniken zur automatischen Gesichtserkennung,
deren Erkennungsgüte zum Teil über der menschlicher Versuchspersonen liegt.
– Jan-Philipp Steghöfer et al. von der Universität Augsburg diskutieren im Beitrag ,,Die Guten, die Bösen und die
Vertrauenswürdigen (Vertrauen im Organic Computing)“ die Nutzung von Vertrauenskonzepten wie funktionale
Korrektheit, Betriebssicherheit, Datensicherheit, Zuverlässigkeit, Glaubwürdigkeit und Gebrauchstauglichkeit für
selbstorganisierende Energienetze und offene Desktop Grid Computing Systeme.
Wir möchten den Autoren für die fristgerechte Fertigstellung und dem Review-Team1 für die kritische Durchsicht der
Beiträge danken.
C. Müller-Schloer
H. Schmeck
T. Ungerer
Literatur
1. Allrutz R, Cap C, Eilers S, Fey D, Haase H, Hochberger C, Karl W, Kolpatzik B, Krebs J, Langhammer F, Lukowicz P, Maehle E, Maas J, Müller-Schloer C,
Riedl R, Schallenberger B, Schanz V, Schmeck H, Schmid D, Schröder-Preikschat W, Ungerer T, Veiser H-O, Wolf L (2003) VDE/ITG/GI – Positionspapier
Organic Computing: Computer- und Systemarchitektur. Gesellschaft für Informatik e.V.
2. Müller-Schloer C, Schmeck H, Ungerer T (2011) Organic Computing: A Paradigm Shift for Complex Systems, Birkhäuser Verlag, http://www.springerlink.
com/content/978-3-0348-0129-4/contents/
1
Die Durchsicht der Beiträge haben dankenswerterweise die folgenden Kolleginnen und Kollegen übernommen: Prof. Dr. J. Branke,
Prof. Dr. U. Brinkschulte, Prof. Dr. J. Hähner, Prof. Dr. A. Herkersdorf, Prof. Dr. W. Karl, Dr. S. Mostaghim, Prof. Dr. G. Mühl, Prof. Dr. B. Sick,
Prof. Dr. J. Teich, Dr. R. Würtz.
Informatik_Spektrum_35_2_2012
73
{ MITTEILUNGEN
DER GESELLSCHAFT FÜR INFORMATIK / 214.
FOLGE
Gesellschaft für Informatik e.V. (GI), Wissenschaftszentrum, Ahrstraße 45, 53175 Bonn, Tel.: 0228/302-145, Fax 0228/302-167,
e-mail: [email protected], Server: http://www.gi.de
Geschäftsführung: Dr. Peter Federer, e-mail: [email protected], Tel.: -145
Stellvertreterin der Geschäftsführung: Cornelia Winter, e-mail: [email protected], Tel.: -147
Stellvertreter des Geschäftsführers: Alexander Rabe, e-mail: [email protected]
Sekretariat: Monika Schulte, e-mail: [email protected], Tel.: -145
Mitgliederwerbung: Ludger Porada, e-mail: [email protected], Tel.: -146
Finanzen/Buchhaltung: Elena Kerkmann, e-mail: [email protected], Tel.: -153, Swetlana Ruppel, e-mail: [email protected], Tel.: -152
Mitgliederverwaltung: Tanja Diegeler, e-mail: [email protected], Tel.: -149, Karola Schmitt, e-mail: [email protected], Tel.: -151
ITK: Christopher Schiller, e-mail: [email protected], Tel.: -156
i
Aus der Geschäftsstelle
Aufruf zur
Nominierung der
GI-Fellows 2012
Alexander Rabe
weiterer
Stellvertreter des
GI-Geschäftsführers
Seit dem 1. Januar 2012 verstärkt
Alexander Rabe, Geschäftsführer
der Deutschen Informatik-Akademie
(DIA), die GI-Geschäftsstelle. Er
wird sich neben der reduzierten
Weiterführung der DIA-Seminare
schwerpunktmäßig um die Gewinnung neuer Mitglieder kümmern.
i
INFORMATIK 2012
in Braunschweig
Wir freuen uns, Sie zur gemeinsamen Jahrestagung der Gesellschaft für Informatik e.V. (GI)
und der Deutschen Gesellschaft
für Medizinische Informatik, Bio-
Tagungsankündigungen
GI-Mitglieder haben auch in
diesem Jahr wieder die Möglichkeit, verdiente Persönlichkeiten
für das GI-Fellowship vorzuschlagen. Die GI zeichnet in jedem Jahr
einige ihrer Mitglieder für besondere
Verdienste um die GI selbst oder um
die Informatik aus. Dieses Programm
haben wir im Jahr 2002 gestartet,
mittlerweile zählen wir gut 70 Fellows in unserer Mitte. Wenn Sie nun
jemanden kennen, der Ihrer Meinung
nach Besonderes für die Informatik
geleistet hat, können Sie ihn oder sie
gemeinsam mit mindestens vier anderen GI-Mitgliedern für das Fellowship
vorschlagen.
Um eine formale Vergleichbarkeit
der Vorschläge zu erreichen, haben
wir eine Nominierungshilfe bereitgestellt, die Sie unter http://www.gi.de/
wir-ueber-uns/personen/fellowship/
zum Herunterladen finden, zusammen mit weiteren Informationen zum
Fellowship und den Auszeichnungskriterien. Einsendeschluss ist der
25. Mai 2012.
Im Frühsommer trifft sich der
Auswahlausschuss, um über die eingegangenen Vorschläge zu beraten und
die Fellows 2012 zu küren. Unsere Bitte
an Sie: Machen Sie regen Gebrauch von
der Möglichkeit, die Ehrungen in Ihrer
Gesellschaft aktiv mitzugestalten. Wir
freuen uns auf viele Vorschläge! Fragen
und Anregungen sowie die Nominierungen senden Sie bitte per E-Mail an
[email protected]. Bei Bedarf steht Ihnen
Cornelia Winter unter 0228-302147 auch
gerne telefonisch zur Verfügung.
metrie und Epidemiologie e.V.
(GMDS) im Jahr 2012 nach Braunschweig einzuladen.
Unter dem Leitthema ,,Was bewegt uns in der/die Zukunft – Neue
Lebenswelten in der Informationsgesellschaft“ werden hier vielfältige
Fragestellungen diskutiert. Neue Lebenswelten spannen einen Bogen
unter anderem von Ambient Assisted
Living Systemen und assistierenden Gesundheitstechnologien bis
zu Lösungen für nachhaltige Mobilität. Auch die Verknüpfung dieser
beiden Megathemen Mobilität und
Gesundheit ist von großer Bedeutung
für unsere zukünftige gesellschaftliche Entwicklung. Auf der ersten
gemeinsamen Tagung von GI und
GMDS werden für unsere Gesellschaft
und für die Forschung wichtige zukünftige Entwicklungen diskutiert;
es wird aber auch ein Blick auf die
Entwicklungen der vergangenen Jahrzehnte geworfen. Wir erwarten einen
persönlichen Austausch, anregende
Diskussionen und individuelle Gespräche und würden uns freuen, Sie
in Braunschweig begrüßen zu dürfen.
www.informatik2012.de
DOI 10.1007/s00287-012-0605-8
Die Fortsetzung der Mitteilungen folgt auf Seite 160.
74
Informatik_Spektrum_35_2_2012
{ MITTEILUNGEN
DER GESELLSCHAFT FÜR INFORMATIK / 214.
Aus Vorstand
und Präsidium
Ergebnisse der
Präsidiumssitzung
vom 2./3. Februar 2012
Mitglieder des erweiterten Vorstands
für die Amtsperiode 2012–2013:
– Prof. Dr. Univ.-Prof. Dr.-Ing. Dr.
techn. Dieter W. Fellner
– Dr. Johannes Helbig
– Prof. Dr. Wolfgang Karl
Bestätigungen und Benennungen:
– Prof. Dr. Ralf Hofestädt als Sprecher des FB ,,Informatik in den
Lebenswissenschaften“
– Axel Kern, Köln als Vertreter des
German Chapter of the ACM
– als LNI-Herausgeber:
– Prof. Dr. Maria Wimmer (FB RVI)
– Prof. Dr. Ingo Timm (FB KI)
– Prof. Dr. Karin Vosseberg (FB IUG)
– Prof. Dr. Steffen Hölldobler
(Dissertations)
Bestätigung als Präsidiumsverantwortliche:
– Hauptherausgeber für das ,Informatik Spektrum‘: Prof. Dr. Dr.-Ing.
Arndt Bode
– INFORMATIK 2012: Prof. Dr. Lars
Wolf
– ,Past President‘: Prof. Dr. Stefan
Jähnichen
– Fachinformationszentrum FIZ
Karlsruhe: Gesellschafterversammlung und Aufsichtsrat: Prof. Dr. Dr.
h.c. Wolffried Stucky
– GI-Dissertationspreis: Prof. Steffen
Hölldobler
– Max-Planck-Gesellschaft (MPG):
Prof. Dr. Bernhard Thalheim
– Vertreter im Stiftungsrat der
Stiftung ,,Werner-von-Siemens-
160
Informatik_Spektrum_35_2_2012
Ring“: Prof. Dr. Dr. h.c. Heinz G.
Schwärtzel
– Anerkennungsausschuss Medizinischer Informatikerinnen und
Informatiker: Prof. Dr. Jürgen
Stausberg
– GI-Vertretung im DIN Normungsausschuss Informationstechnik
(NIA): Dr. Hans von Sommerfeld
Ständige Gäste im Präsidium:
– Vorsitzender des Fakultätentages
Informatik
– Vorsitzender des Fachbereichstages
Informatik
– Sprecher des Querschnittsfachausschusses ,,Modellierung“
(QFAM)
Die Präsidiumsarbeitskreise ,,Datenschutz und IT-Sicherheit“ und ,,Bild
der Informatik“ werden um ein Jahr
verlängert.
Unsere
Fördermitglieder
Accso -- Accelerated Solutions GmbH
AKAD Hochschule Stuttgart
Akademie für Datenverarbeitung
AKDB -- Anstalt für Kommunale Datenverarbeitung in Bayern
A·S Technology Consulting GmbH
Aschert & Bohrmann GmbH
ask-Innovative Visualisierungslösungen
GmbH
Avius Software GmbH
Barmenia Allgemeine Versicherungs AG
Bayer Business Services GmbH
Bayerisches Landesamt für Statistik &
Datenverarbeitung
BearingPoint GmbH
Berufsakademie Thüringen
Berufsfachschule für IT-Berufe
BESIT e.V.
Beuth Hochschule für Technik Berlin
Bildungszentrum für informationsverarbeitende Berufe e.V.
BMW AG
Büren & Partner
Bundesamt für Sicherheit in der
Informationstechnik
Bundesanstalt für Landwirtschaft und
Ernährung
Bundesverband IT-Mittelstand e.V. (BITMi)
BwFuhrpark Service GmbH
C1 SetCon GmbH
Capgemini sd&m AG
Carl Hanser Verlag GmbH & Co. KG
Carl-Cranz-Gesellschaft e.V.
CAST e. V.
ckc ag
Collogia Unternehmensberatung AG
come2trace
Comma Soft AG
Commerzbank AG
CommitWork GmbH
Con.ECT Eventmanagement GmbH
Darmstadt
Stuttgart
Böblingen
München
Frankfurt am Main
Köln
Darmstadt
Bad Soden
Wuppertal
Leverkusen
München
Hamburg
Gera
Plattling
Nürnberg
Berlin
Paderborn
München
Nürnberg
Bonn
Bonn
Aachen
Troisdorf
München
München
München
Weßling
Darmstadt
Braunschweig
Köln
Mainz
Bonn
Frankfurt
Dortmund
Wien
FOLGE (FORTSETZUNG)
con terra GmbH
COR&FJA Deutschland GmbH
CoreMedia AG
Dataport
DAVID Software GmbH
Der Hessische Datenschutzbeauftragte
DESY Deutsches Elektronen Synchroton
Deutsche Bank AG
Deutsche Börse AG
Deutsches Forschungszentrum
Dexia Kommunalbank Deutschland AG
DHBW Heidenheim
Diamant Software GmbH & Co. KG
DLGI Dienstleistungsgesellschaft für
Informatik mbH
Duale Hochschule Baden-Württemberg
Duale Hochschule Baden-Württemberg
dycept GmbH
EDUARD-RHEIN-STIFTUNG
edv-anwendungsberatung zühlke & bieker
gmbH
Eridea AG
EUROFORUM Deutschland SE
Europäische Bildungsakademie
eXept AG
Fachbereich DCSM
Fachhochschule Aachen
Fachhochschule Amberg-Weiden
Fachhochschule Brandenburg
Fachhochschule der Wirtschaft Hannover
Fachhochschule Dortmund
Fachhochschule Gelsenkirchen
Fachhochschule Hannover
Fachhochschule Kempten
Fachhochschule Münster
Fachhochschule Niederrhein
Fachhochschule Schmalkalden
Fachhochschule Stralsund
Fachhochschule Wedel
Fakultät für Informatik der TU München
FIZ Karlsruhe Gesellschaft für wissenschaftlich technische Informationen mbh
Forschungszentrum Informatik
Forschungszentrum Jülich GmbH
Fraunhofer IESE
Fraunhofer IITB
Fraunhofer Institut IAO
Fraunhofer-Gesellschaft e.V.
Fraunhofer-Gesellschaft e.V.
Fraunhofer-Gesellschaft zur Förderung
der Angewandten Forschung e.V.
Fraunhofer-Institut-IPA
Freie Universität Bozen
Friedrich-Schiller-Universität
G.E.M. Team Solutions GbR
Generali Deutschland Informatik Services
GmbH
GeoForschungsZentrum Potsdam
Georg-August-Universität Göttingen
Gesellschaft der Förderer und Freunde
der HS Würzburg-Schweinfurt e.V.
-- IWInet
Gesellschaft für Datenschutz und
Datensicherheit e.V.
Gesellschaft für Systementwicklung
und Datenverarbeitung
Gesellschaft für wissenschaftliche
Datenverarbeitung mbH
GESIS
Gimtec GmbH
Goerigk Informationstechnologie GmbH
GORBIT GmbH
Hasso-Plattner-Institut für SW-Systemtechnik GmbH
HDI-Gerling Leben Betriebsservice GmbH
Heinz Nixdorf Institut
Helmholtz-Zentrum Berlin für Materialien
Helmholtz-Zentrum Geesthacht
Helmut Schmidt Universität
Hessische Zentrale für Datenverarbeitung
HIS Hochschul-Informationssysteme GmbH
HMS Analytical Software GmbH
HNF Heinz Nixdorf MuseumsForum GmbH
Hochschule Anhalt
Hochschule Bremerhaven
Hochschule der Medien
Hochschule Esslingen
Hochschule Fulda
Hochschule Furtwangen
Hochschule Ruhr West
Hochschule Weserbergland
Hochschule Wismar
HPA Hamburg Port Authority AöR
Humboldt Universität Berlin
DOI 10.1007/s00287-012-0603-x
Münster
München
Hamburg
Altenholz
Braunschweig
Wiesbaden
Hamburg
Eschborn
Frankfurt
Kaiserslautern
Berlin
Heidenheim
Bielefeld
Bonn
Stuttgart
Karlsruhe
Kornwestheim
Mayen
Recklinghausen
Bad Aibling
Düsseldorf
Berlin
Bönnigheim
Wiesbaden
Aachen
Amberg
Brandenburg
Hannover
Dortmund
Gelsenkirchen
Hannover
Kempten
Münster
Krefeld
Schmalkalden
Stralsund
Wedel
Garching
EggensteinLeopoldshafen
Karlsruhe
Jülich
Kaiserslautern
Karlsruhe
Stuttgart
Berlin
München
Wachtberg
Stuttgart
Bozen
Jena
Neustadt
Aachen
Potsdam
Göttingen
Würzburg
Bonn
Erlangen
Göttingen
Bonn
Saarbrücken
Bad Nauheim
Bergisch
Gladbach
Potsdam
Köln
Paderborn
Berlin
Geesthacht
Hamburg
Wiesbaden
Hannover
Heidelberg
Paderborn
Köthen
Bremerhaven
Stuttgart
Esslingen
Fulda
VillingenSchwenningen
Bottrop
Hameln
Wismar
Hamburg
Berlin
IBM Deutschland GmbH
ICS AG
imbus AG
Infodas Gesellschaft für Systementwicklung
Informatikzentrum Landesverwaltung BW
Information und Technik NRW
Information Works GmbH
innoQ Deutschland GmbH
innovas Innovative Anwendungssysteme
Institut für Ökonometrie
Institut für Rundfunktechnik Gmbh
Institutsverbund Informatik
it-and-finance
IT-Forum Rhein-Neckar e. V.
IT Science Center Rügen gGbmH
IT Service Omikron GmbH (ITSO)
ITech Progress GmbH
iteratec GmbH
itestra GmbH
IVU Traffic Technologies AG
jambit GmbH
Karlsruher Institut für Technologie -- KIT
Karlsruher Institut für Technologie (KIT)
Kern AG
Kölsch & Altmann GmbH
KRIWAN Testzentrum GmbH & Co. KG
Landesinstitut für Schulentwicklung
Login-IT GmbH & Co. KG
MaibornWolff et al. GmbH
Max-Planck-Institut für Informatik
Rechenzentrum Garching der Max Planck
Gesellschaft (RZG)
McKinsey & Company
mehrwert intermediale kommunikation
GmbH
Mio Soft Deutschland GmbH
msg systems ag
Neubert Consulting GmbH
NORDAKADEMIE
Normfall GmbH
NovaTec GmbH
OFFIS e.V.
Oldenbourg Verlag GmbH
Opitz Müller und Partner GbR
Ostfalia Hochschule für angewandte
Wissenschaften
Otto-von-Guericke-Universität Magdeburg
Paranor AG
Patentanwältin Dr. Claudia Schwarz
Patzschke+Rasp Software AG
Pearson Education Deutschland GmbH
perbit Software GmbH
PPI AG Informationstechnologie
Private FernFachhochschule Darmstadt
PROJECT CONSULT GmbH
PROMATIS software GmbH
QAware GmbH
Quality Team Deutschland GmbH
Qualmity
Rechenzentrum der Finanzverwaltung
NRW
Rheinmetall Defence Electronics GmbH
RISA Sicherheitsanalysen GmbH
Robert Bosch GmbH
RWTH Aachen
RWTH Aachen (IMA)
SAP AG
Schumacher Informatik AG
SCM Software-Labor GmbH
Seeliger & Co. GmbH
sepp.med gmbh
Siemens AG
SIZ.Informationszentrum der Sparkassenorganisation GmbH
SMO GmbH
SNP Schneider-Neureither & Partner AG
Software AG
Softwarekontor GmbH
SOPTIM AG
SPIQ e.V.
Springer-Verlag GmbH
SQLan Gesellschaft für Informations- und
Netzwerksysteme mbH
SRH Hochschule Heidelberg
Staats- u. Universitätsbibliothek
Staufenbiel Institut für Studien- und
Berufsplanung
Stiftung Alfred-Wegener-Institut
supERP-Software UG
SYRACOM AG
SYSLAB.COM GmbH
Technische Akademie Ahaus
Technische Akademie Esslingen
TU Berlin
Technische Universität Berlin
Ehningen
Stuttgart
Möhrendorf
Köln
Stuttgart
Düsseldorf
Köln
Ratingen
Hamburg
Bonn
München
Stuttgart
München
Ludwigshafen
Putbus
Berlin
Ludwigshafen
Unterhaching
München
Berlin
München
EggensteinLeopoldshafen
Karlsruhe
Freiburg
München
Forchtenberg
Stuttgart
Kirchheimbolanden
München
Saarbrücken
Garching
Düsseldorf
Köln
Möhnesee
Ismaning
Fürth
Elmshorn
München
LeinfeldenEchterdingen
Oldenburg
München
Berlin
Wolfenbüttel
Magdeburg
Wahlendorf
München
Wiesbaden
München
Altenberge
Hamburg
Darmstadt
Hamburg
Ettlingen
München
München
Berlin
Düsseldorf
Bremen
Berlin
Stuttgart
Aachen
Aachen
Walldorf
Gerlingen
Kehl
Eichenau
Röttenbach
München
Bonn
Karlsruhe
Heidelberg
Darmstadt
Ludwigshafen
Aachen
Freiburg
Heidelberg
Wiesbaden
Heidelberg
Bremen
Köln
Bremerhaven
Hofbieber
Wiesbaden
München
Ahaus
Ostfildern
Berlin
Berlin
Technische Universität Chemnitz
Technische Universität Dresden
Technische Universität in Brno
Tekko Informationssysteme GmbH
& Co.KG
TELENET GmbH
The MathWorks GmbH
Tideum Deutschland AG
Transaction Software GmbH
TÜV Informationstechnik GmbH
UGE UkrGermanEnterprise GmbH
Unister Holding GmbH
Uni- und Stadtbibliothek -Medienbearbeitung
Universität Bremen
Universität der Bundeswehr München
Universität des Saarlandes
Universität Dortmund
Universität Duisburg-Essen
Universität Erlangen-Nürnberg
Universität Hamburg
Universität Hamburg
Universität Koblenz-Landau
Universität Leipzig
Universität Lüneburg
Universität Mannheim
Universität Osnabrück
Universität Paderborn
Universität Rostock
Universitätsbibliothek Bamberg
TU Dortmund
velian GmbH
Verband der Vereine Creditreform e. V.
Versant GmbH
V-Research GmbH
W.-Schickard-Institut der Universität
Tübingen
Werum Software & Systems AG
Westfälische Wilhelms-Universität
WidasConcepts GmbH
Wifa Consulting & Education
-Wifa-Gruppe- GmbH
Wirtschaftsuniversität Wien
Wirtschaftswissenschaftliches Seminar
der Universität Tübingen
zisa consulting GmbH
Chemnitz
Dresden
Brno
Kiel
München
Ismaning
Kaarst
München
Essen
Frankfurt
Leipzig
Köln
Bremen
Neubiberg
Saarbrücken
Dortmund
Essen
Erlangen
Hamburg
Hamburg
Koblenz
Leipzig
Lüneburg
Mannheim
Osnabrück
Paderborn
Rostock
Bamberg
Dortmund
Braunschweig
Neuss
Hamburg
Dornbirn
Tübingen
Lüneburg
Münster
Wimsheim
Bad Honnef -Rhöndorf
Wien
Tübingen
Berlin
Aus den
GI-Gliederungen
Ehrungen im Fachausschuss
Echtzeitsysteme/Realtime
In den vergangenen Jahren hat der
Fachausschuss bereits Preise für
Abschlussarbeiten besonderer Qualität vergeben. Auch 2011 gab es
wieder zwei Preisträger des Wettbewerbs. Darüber hinaus gab es für
2011 die Ernennung eines Ehrenmitgliedes des Fachausschusses und es
wurde ein ,,Best Paper“ Award des
Springer-Verlages vergeben.
Die jeweils mit 500 Euro dotierten Preise für besonders gelungene
Abschlussarbeiten wurden an Frank
Engelhardt und Till Fischer vergeben. Herr Engelhardt beschäftigte
sich in seiner Arbeit mit der ,,Umsetzung eines Online-SLAM-Verfahrens
auf der Roboterplattform VolksbotLab“. Der Titel der Arbeit von Herrn
Fischer lautet: ,,Entwurf eines FPGACores zur Simulationsbeschleuigung
zeitkontinuierlicher Modelle im
HiL-Kontext“.
Herrn Prof. Dr. Helmut Rzehak (siehe Abb. 1) wurde als einem
der Männer der ersten Stunde des
Fachausschusses durch Prof. Dr. Halang eine Ehrenmitgliedschaft im
Fachausschuss verliehen.
Den Preis für den besten Tagungsbeitrag durften Alexander Hug
und Andreas Stahlhofen für ihren
Beitrag ,,Echtzeitsysteme in Informatikunterricht und Ausbildung“
entgegen nehmen.
Für das kommende Jahr ist
ebenfalls ein Wettbewerb zur Nachwuchsförderung geplant. Außer
über eine Notiz in der InformatikSpektrum können sich Interessierte
auf der Homepage des Fachausschusses www.real-time.de über die
Ausschreibung informieren.
Neues aus der GI-Fachgruppe
Management von Informationssicherheit (SECMGT)
Die GI-Fachgruppe Management von
Informationssicherheit (SECMGT)
hat durch ihre Fusion mit der
zum Jahreswechsel aufgelösten
GI-Fachgruppe KRITIS zu den bestehenden 252 Mitgliedern weitere
40 Mitglieder hinzugewonnen und ist
nunmehr die mitgliederstärkste Fachgruppe im Fachbereich Sicherheit. Im
zurückliegenden Jahr konnten mit
ansprechenden Veranstaltungen zu
aktuellen Praxisthemen zahlreiche
Interessenten für Aktivitäten der GI
gewonnen werden.
Innerhalb der GI-FG SECMGT
besteht seit September 2011
ein Arbeitskreis zu kritischen
Informations- und Kommunikationsinfrastrukturen (AK KRITIS),
der einerseits die Themen aus der
zum Jahreswechsel aufgelösten FG
KRITIS fortführt und andererseits
Informatik_Spektrum_35_2_2012
161
{ MITTEILUNGEN
DER GESELLSCHAFT FÜR INFORMATIK / 214.
FOLGE (FORTSETZUNG)
9. ITG/GI/GMM – Workshop
Cyber-Physical Systems –
Enabling Multi-Nature
Systems (CPMNS 2012)
Abb. 1 Prof. Dr. Zöbel, Prof. Dr. Rzehak und Prof. Dr. Halang (von links)
diese Themen unter dem Fokus des
Sicherheitsmanagements betrachtet.
Details zum AK KRITIS können der
Web-Seite unter http://www.secmgt.
de/kritis.html entnommen werden.
Die FG SECMGT hat sich im zurückliegenden Jahr zweier weiterer
,,heißer Eisen“ angenommen: Zum
Einen, wie ungewollter Informationsfluss detektiert und vorzugsweise verhindert werden kann, im
Workshop über Data Leakage (48 Teilnehmer) und zum Anderen, welche
Rolle der Informationssicherheit
beim Outsourcing zukommt, im
Workshop über Outsourcing und
Vendor Security (45 Teilnehmer).
Workshops der FG SECMGT
berechtigen zum Erwerb kontinuierlicher Fortbildungspunkte
(CPE) für alle einschlägigen
162
Informatik_Spektrum_35_2_2012
Personalzertifizierungen zur Informationssicherheit. Die Unterlagen
zu den Workshops der FG SECMGT
können unter http://www.secmgt.
de/workshops/ abgerufen werden.
Die FG SECMGT hat 2011 zusammen mit der GI-FG PET auf der
GI-Jahrestagung eine neue Reihe
ins Leben gerufen: Workshops mit
besonderem Anwendungsfokus zu
Sicherheitsmanagement und Datenschutz in Anwendung und Praxis.
Dies wird 2012 durch eine Zusammenarbeit mit der AG DGI der GMDS
flankiert. Im Bereich der kritischen
Informations- und Kommunikationsinfrastrukturen wurde 2011 mit dem
FIfF kooperiert.
Ferner wurde 2011 der Webauftritt der FG SECMGT neu gestaltet:
http://www.secmgt.de.
Vom 17.–18. April 2012 findet in Dresden der 9. Workshop Cyber-Physical
Systems – Enabling Multi-Nature
Systems statt. Dieser führt die
langjährige Themenstellung des
Workshops Multi-Nature Systems
fort und wird von der Fachgruppe
RSS-METHOD ,,Allgemeine Entwurfsmethodik und Unterstützung
von Entwurfsprozessen für Schaltungen und Systeme“ in Kooperation
mit dem Institutsteil Entwurfsautomatisierung EAS des Fraunhofer IIS
organisiert.
Im Mittelpunkt des Interesses
steht der domänenübergreifende
Entwurf von heterogenen eingebetteten Systemen mit gemeinsamen
mikro- und nanoelektronischen und
nichtelektronischen Komponenten –
wie z. B. optischen, mechanischen,
thermischen, fluidischen, akustischen oder biologischen Systemen.
Die Thematik umfasst sowohl die
Sicht auf Cyber-Physical Systems
als Ganzes (Spezifikation, Modellierung, Entwurf, Beispiele) als auch
deren Systemkomponenten (Sensoren, Aktoren, spezielle elektronische
Komponenten) und ihre Anwendung.
IT-Projektmanagement
2012+ im Spagat zwischen
Industrialisierung und
Agilität? 11.–12. Mai 2012,
Glashütten bei Frankfurt am
Main
Die Diskussion im IT-Projektmanagement wird von zwei Ansätzen
geprägt: Agiles Projektmanagement und Industrialisierung der
Softwareentwicklung. Agiles Projektmanagement hat seinen Ursprung
in Softwareentwicklungsprojekten und hat sich dort in der Praxis
bewährt. Agile Vorgehensweisen,
Scrum, SW-Kanban & Co. entfalten
ihr größtes Potenzial offenbar bei
komplexen Entwicklungsvorhaben,
also in einem dynamischen Kontext
mit vielen Überraschungen. Dem
gegenüber steht die Industrialisierung der Softwareentwicklung, die
durch standardisierte und modularisierte Prozesse die Grundlagen
zur stärkeren Automatisierung,
Wiederverwendung und Variationen der Spezialisierung in den
Entwicklungsprojekten schafft. Im
Rahmen der interPM 2012 diskutieren Wissenschaftler und Praktiker
darüber, welche Erkenntnisse zu
agilen und industrialisierten Ansätze des IT-Projektmanagements
vorliegen und wie sich der Spagat
zwischen den beiden Grundlinien
schließen lässt. Veranstaltungsprogramm und Anmeldung unter
www.interpm.de. Ansprechpartner:
Prof. Dr. Martin Engstler (Hochschule
der Medien Stuttgart), stv. Sprecher
der Fachgruppe WI-PM
,,Mensch & Computer 2012“
in Konstanz
Der Touchscreen des ultraleichten
Tablet-PCs versteht unsere Gestensprache, das mobile Smartphone
begleitet uns auf Geschäftsreisen, das
berührungsempfindliche Wanddisplay erschließt neue Raumkonzepte
im Museum und die 3D-Kamera trägt
unsere Bewegungen ins Videospiel
hinein. Unsere zunehmend feingliedrigere Interaktion mit digitalen
,,Informationswelten“ verändert unser Kommunikationsverhalten und
nicht zuletzt die Art und Weise, wie
wir Entscheidungen treffen.
Die Fachtagung ,,Mensch & Computer 2012“ in Konstanz nimmt diese
technologischen Entwicklungen
zum Anlass, um unter dem Motto
,,interaktiv informiert – allgegenwärtig & allumfassend!?“ Beiträge
von Forschern der Fachdisziplin
,,Mensch-Computer-Interaktion“ zur
Entwicklung von neuen Interaktionskonzepten vorzustellen und kritisch
zu diskutieren.
Im Rahmen von Fachvorträgen,
Workshops, Podiumsdiskussionen,
Tutorien, Ausstellungen und Demonstrationen werden vielfältige
Gelegenheiten bestehen, um den
aktuellen Stand der Forschung zu
präsentieren und mit Kollegen aus der
Forschung und Praxis zu diskutieren.
Der Fachbereich Mensch-Computer
Interaktion der Gesellschaft für Informatik und die German UPA e.V. laden
zur Tagung ein.
Interaktion-Demos bzw. -Kurzbeiträge werden bis 13. Mai 2012 erbeten und Workshop-Papiere und
Tutorien können bis zum 17. Juni
2012 eingereicht werden. Interdisziplinäre Arbeiten und Beiträge mit
einem Bezug zum Tagungsthema ,,interaktiv informiert – allgegenwärtig
& allumfassend!?“ sind besonders erwünscht. Nähere Informationen unter
http://hci.uni-konstanz.de/mc2012/.
IFIP World Computer
Congress 2012 (WCC2012)
in Amsterdam,
24.–26. September 2012
Der 22. WCC (http://www.wcc-2012.
org) wird dieses Jahr in Amsterdam
stattfinden, und dies mit einem völlig neuartigem Konzept. Die Ngi, die
Niederländische Informatikgesellschaft lädt zahlreiche internationale
Referenten nach Amsterdam ein,
um eine Konferenz auf höchstem
Niveau zu organisieren. Dabei werden zentrale Fragestellungen in und
um die Informations- und Kommunikationstechnik (IKT) behandelt.
IKT ist heute allgegenwärtig. Als Beispiele sollen hier nur die Bereiche
Soziale Medien, Cloud Computing,
Mobile Plattformen, große Datenbanken, Unternehmens-Software oder
die Informatisierung der Technik
genannt werden. Das Internet ermöglicht weltweite Geschäftsprozesse, es
ermöglicht eine kollaborative Kommunikation und gewährt Zugang zu
ungeahnten Informationsmengen.
Der uralte Menschheitstraum von der
Überwindung von Zeit und Raum
scheint Wirklichkeit geworden zu
sein. Heute hängen die Lebensumstände ganz wesentlich von der IKT
ab. Nahezu alle Berufstätigen müssen daher den Erfordernissen einer
modernen IKT-geprägten Berufswelt
gerecht werden. Die Teilnehmer werden nicht nur bei den Diskussionen
während des Kongresses eingebunden sein, sondern bereits im Vorfeld
des WCC2012 über Webseiten und
andere sozialen Medien Themen diskutieren können. Es ist geplant, die
WCC2012 mit einer Agenda für die
nächsten 10 Jahre abzuschließen.
Herbstschule 2012 der
Fachgruppe Information
Retrieval
In der bewährten Atmosphäre des
Leibniz-Zentrums für Informatik
Schloss Dagstuhl findet vom 1.10.
bis zum 5.10.2012 wieder die Herbstschule Information Retrieval der
GI-Fachgruppe Information Retrieval (FG-IR) statt. Die Vorträge der
Herbstschule werden ein breites thematisches Spektrum des Information
Retrieval abdecken und richten sich
an Interessenten aus Wissenschaft
und Praxis: Studierende, Wissenschaftler, Anwender und Mitarbeiter
von Institutionen und Unternehmen
aus den Bereichen Suchmaschinen,
Information Retrieval, Datenbanken,
Informationssysteme sind herzlich
willkommen.
Durch die Kombination grundlegender Themen und aktueller
Forschungsfragen des Information
Retrieval ist die Herbstschule besonders auch für junge Wissenschaftler
geeignet, die ihre Kenntnisse z. B.
Informatik_Spektrum_35_2_2012
163
{ MITTEILUNGEN
DER GESELLSCHAFT FÜR INFORMATIK / 214.
in einem frühen Stadium ihres
Promotionsvorhabens oder im
Masterstudium fundieren und verbreitern möchten. Praktikern bietet
die Herbstschule einen fundierten
Überblick über Methoden, Konzepte
und aktuelle Forschungsfragen des
Information Retrieval mit hohem
Praxisnutzen.
Weitere Informationen finden Sie
unter http://tinyurl.com/HSIR2012.
Dort können Sie sich auch registrieren, um über die Themen und die
Dozenten sowie die Modalitäten der
Anmeldung informiert zu werden.
(Andreas Henrich)
Tagungsberichte
Workshop der Fachgruppe
Information Retrieval im
Rahmen der LWA 2011
Der schon traditionelle Herbstworkshop der Fachgruppe Information
Retrieval (IR) fand vom 28.–30.
November 2011 im Rahmen der
Konferenz ,,Lernen, Wissen & Adaptivität“ (LWA) 2011 in Magdeburg
statt. Traditionell bot die LWA wieder
ein Dach für die Workshops verschiedener Fachgruppen innerhalb der
Gesellschaft für Informatik. Ziel des
IR-Workshops war es, ein Forum
zur wissenschaftlichen Diskussion
und zum Austausch neuer Ideen zu
schaffen. Der Workshop richtete sich
daher gezielt auch an Nachwuchswissenschaftler und Teilnehmer aus
der Industrie. Dementsprechend
wurden interessante Vorträge zu aktuellen Themen aus dem Information
Retrieval wie Informationsextraktion, Benutzermodellierung,
Suche in sozialen Netzen, Interaktives Information Retrieval und
Anwendungsszenarien geboten. Gemeinsamen Vortragssitzungen mit
den Fachgruppen Knowledge Discovery, Data Mining und Maschinelles
164
Informatik_Spektrum_35_2_2012
Lernen sowie Wissensmanagement
waren eine willkommene Gelegenheit, Arbeiten auch über die
Grenzen der eigenen Fachgruppe
hinaus zu präsentieren und zu
diskutieren.
Drei Keynotes rundeten die
LWA in diesem Jahr ab: Daniel Keim
(Universität Konstanz) sprach über
Herausforderungen und Anwendungen von Visual Analytics, Christian
Müller-Schloer (Universität Hannover) stellte ,,Organic Computing“
vor, und Harald Sack (HPI Potsdam)
sprach über ,,SeMEX – Enabling Exploratory Video Search by Semantic
Video Analysis“.
BIOSIG 2011
am 8.–9. September 2011
in Darmstadt
Durch die zunehmende Verbreitung
von elektronischen Reisepässen und
Bürgerkarten haben zahlreiche Bürger am Registrierungsprozess von
biometrischen Merkmalen bereits
aktiv teilgenommen. Biometrische
Verifikation an den Schengen Grenzen mit ePassports und dem Visa
Informationssystem (VIS) werden
zurzeit auf breiter Front eingeführt.
Auf der anderen Seite gibt es noch
zahlreiche ungelöste Forschungsfragen, die noch bearbeitet werden
müssen. Die BIOSIG 2011 Konferenz
betrachtete daher aktuelle Themenbereiche wie Biometrische Standards
und Interoperabilität, multimodale Biometrie, Sicherheitsanalysen
biometrischer Komponenten und
Systeme, User Interface Design
biometrischer Systeme, PerformanceMessungen, Best Practices, neuartige
Anwendungen, ethische, rechtliche und sozio-technische Aspekte
und Biometrie für die öffentliche
Verwaltung, sowie weitere Themenbereiche. In einer Poster Session
wurden weitere interessante Forschungsarbeiten vorgestellt. Darüber
FOLGE (FORTSETZUNG)
hinaus fand die Mitgliederversammlung der Fachgruppe BIOSIG statt,
in der das neue Leitungsgremium
gewählt wurde. Den Abschluss des
ersten Tages bildete eine gemeinsame Abendveranstaltung mit einem
Barbecue in gemütlicher Runde. Die
Veranstalter konnten auf der Konferenz mehr als 100 Teilnehmer aus
15 Ländern, aus Industrie, Hochschule, Forschung und Wissenschaft
willkommen heißen. Die Leitung
übernahmen Christoph Busch und
Arslan Brömme.
Die Veranstaltung wurde gemeinsam organisiert von CAST, BSI, JRC,
EBF, TeleTrusT, CASED und der Fachgruppe BIOSIG. Die Vorträge sind auf
www.cast-ev.de. Auch in 2012 wird
die Fachgruppe BIOSIG in Kooperation mit IEEE eine internationale
Biometrie-Konferenz durchführen.
Der CfP ist zu finden unter: http://
www.biosig.org/biosig2012.
Mehr Informationen unter:
http://www.cast-forum.de/files/
download/workshops/biosig-bericht111220.pdf.
(Heiko Roßnagel)
perspeGKtive 2011
am 7. September 2011
in Darmstadt
Bereits heute werden vielfältige
Informationstechnologien im
Gesundheitswesen eingesetzt. Zukünftige organisationsübergreifende
Telematikinfrastrukturen sowie technologische Innovationen werfen
ihre Schatten voraus. Aufgrund des
sehr hohen Schutzbedarfs der im
Gesundheitswesen verarbeiteten
Daten und des Grundsatzes der informationellen Selbstbestimmung
der Patienten kommt den Aspekten des Datenschutzes, der Beweisund IT-Sicherheit eine essenzielle
Bedeutung zu. Vor diesem Hintergrund wurden auf der perspeGKtive
2011 Konferenz Sicherheits- und Da-
tenschutzaspekte der heute und in
Zukunft im Gesundheitswesen genutzten Informationstechnologien
diskutiert und innovative Lösungsansätze für die Gesundheitswirtschaft
von morgen aufgezeigt. Hierbei
reichte die Bandbreite von der elektronischen Überweisung über das
sichere Identitätsmanagement bis hin
zu Aspekten des Cloud Computing
im Gesundheitswesen. Darüber hinaus wurden aktuelle Informationen
präsentiert und interessante Diskussionen zum rechtssicheren Scannen
und der beweiskräftigen Aufbewahrung von Krankenunterlagen und
-akten sowie zur eGK-Sicherheit
geführt.
Die Veranstalter konnten auf der
Konferenz mehr als 70 Teilnehmer
aus Wirtschaft, Wissenschaft und
Verwaltung willkommen heißen. Die
Leitung der Konferenz übernahmen
Arslan Brömme, Torsten Eymann,
Detlef Hühnlein, Heiko Roßnagel und
Paul Schmücker.
In der Proceedingsreihe der
GI-Edition ,,Lecture Notes in Informatics“ sind neu erschienen:
P-188: DeLFI 2011 – Die 9. e-Learning
Fachtagung Informatik, Holger Rohland, Andrea Kienle,
Steffen Friedrich (Hrsg.)
P-189: Informatik in Schule und Beruf – INFOS 2011 – 14. GIFachtagung Informatik und
Schule, Marco Thomas (Hrsg.)
P-190: Enterprise Modelling and Information Systems Architectures (EMISA), Markus Nüttgens, Oliver Thomas, Barbara
Weber
P-191: BIOSIG 2011, Arslan Brömme,
Christoph Busch (Eds.)
P-192: INFORMATIK 2011, HansUlrich Heiß, Peter Pepper,
Holger Schlingloff, Jörg Schneider (Hrsg.)
P-193: IMDM 2011, Wolfgang Lehner,
Gunther Piller (Hrsg.)
P-187: 4. DFN-Forum Kommunikationstechnologien, Paul Müller,
Bernhard Neumair, Gabi Dreo
Rodosek (Hrsg.)
Unter www.gi.de/service/publikationen
finden Sie alle Bände der
Proceedings-, Seminars, Thematicsund Dissertations-Reihe.
Die Veranstaltung wurde gemeinsam von der GI und der GMDS unter
Beteiligung verschiedener Fach- und
Arbeitsgruppen veranstaltet. Die
Veranstaltung wurde unterstützt
von: BVMI, BSI, BMG, BITKOM,
CASED, CAST, CCESigG, DGBMT,
DGG, DGTelemed, GDD, TMF, TeleTrusT, VHitG, VOI. Die Vorträge und
Beiträge sowie weitere Informationen sind auf der Webseite verfügbar
(www.perspegktive.de).
(Heiko Roßnagel)
LNINeuerscheinungen
Informatik_Spektrum_35_2_2012
165
{ GI-VERANSTALTUNGSKALENDER
27.03.–28.03.2012 – Düsseldorf
Qualitätsmanagement und Vorgehensmodelle: 19. Workshop der
Fachgruppe Vorgehensmodelle im
Fachgebiet Wirtschaftsinformatik
GI-VM 2012
http://vorgehensmodelle.de
31.03.2012 – Tallinn/Estonia
9th International Workshop on
Formal Engineering Approaches
to Software Components and
Architectures
FESCA 2012
http://fesca.ipd.kit.edu/fesca2012/
11.05.–12.05.2012 – Glashütten
IT-Projektmanagement 2012+
interPM 2012
http://www.interpm.de
14.05.–18.05.2012 – Paderborn
10th International Symposium on
Modeling and Optimization in Mobile, Ad Hoc, and Wireless Networks
WiOpt 2012
http://www.wi-opt.org/
04.04.–05.04.2012 – Vienna/Austria
S-BPM ONE 2012
http://www.s-bpm-one.org/
index.php?id=conf2012
21.05.–22.05.2012 – Regensburg
5. DFN-Forum Kommunikationstechnologien – Verteilte Systeme im
Wissenschaftsbereich
http://dfn2012.uni-regensburg.de
12.04.–15.04.2012 – Berlin
11th international Conference on
Cognitive Modelling
ICCM 2012
http://www.iccm2012.com/
21.05.–23.05.2012 – Darmstadt
15th IACR International Conference
on Practice and Theory of PublicKey Cryptography
PKC 2012
http://pkc2012.cased.de
17.04.–18.04.2012 – Dresden
9. ITG/GI/GMM Workshop CyberPhysical Systems / Enabling MultiNature Systems
CPMNS 2012
http://www.eas.iis.fraunhofer.de/de/
veranstaltungen/cpmns.html
23.04.–24.04.2012 – Saarbrücken
,,Future e-Learning Technologies in
Medicine and Healthcare“ – 16. CBT
AG Workshop der GMDS
http://www.cbt-ag.de
26.04.–27.04.2012 – Berlin
13. Tagung der Fachgruppe BUIS/
4. Berliner BUIS Tage 2012
http://enviroinfo.eu/4-berliner-buistage
02.05.–04.05.2012 – Bad Honnef
14. Workshop SoftwareReengineering
http://www.fg-sre.gi-ev.de/
konferenzen/wsr.html
166
08.05.–09.05.2012 – Hannover
edaWorkshop 2012
http://www.edacentrum.de/
edaworkshop
Informatik_Spektrum_35_2_2012
21.05.–23.05.2012 – Vilnius/
Lithuania
Business Information Systems –
International Conference
BIS 2012
http://bis.kie.ue.poznan.pl/15th_bis/
21.06.–22.06.2012 – Rostock
eLearning Baltics 2012
eLBa 2012
http://www.e-learning-baltics.de/
science/?L=1.
25.06.–26.06.2012 – Düsseldorf
gearconf – Software entwickeln im
Team
http://www.gearconf.com
27.06.–30.06.2012 – Pisa/Italy
Computer Assisted Radiology and
Surgery – 26th International Congress and Exhibition
CARS 2012
http://www.cars-int.org
27.06.–29.06.2012 – Hamburg
Petri Nets 2012
PN 2012
http://www.informatik.uni-hamburg.
de/TGI/events/pn2012/
27.06.–29.06.2012 – Hamburg
12th International Conference on
Application of Concurrency to
System Design
ACSD 2012
http://www.informatik.uni-hamburg.
de/TGI/events/acsd2012/acsd2012.
shtml
11.06.–13.06.2012 – Passau
International Conference on
Program Comprehension
ICPC 2012
http://icpc12.sosy-lab.org/
02.07.–04.07.2012 – St. Dié des
Vosges/France
Digital Cultural Heritage
DCH 2012
http://dch2012.net
17.06.–21.06.2012 – Hamburg
International Supercomputing
Conference
ISC 2012
http://www.isc-events.com/isc12/
16.07.–20.07.2012 – Alsancak, Izmir/
Turkey
1st IEEE International Workshop on
Methods for Establishing Trust with
Open Data
METHOD 2012
http://compsac.cs.iastate.edu/
workshop_details.php?id=51
20.06.–22.06.2012 – München
Smart Grid – Smart Law? Legal
challenges for new and upcoming
eBusiness models
IFCLA 2012
http://www.ifcla2012.com
DOI 10.1007/s00287-012-0604-9
16.07.–20.07.2012 – Berlin
6th ACM International Conference
on Distributed Event-Based Systems
http://www.csw.inf.fu-berlin.de/
debs2012/
05.08.–10.08.2012 – Bremen
The Tenth IFIP International
Working Conference on Information Technology in Educational
Management
ITEM 2012
http://www.informatik.uni-bremen.
de/agim/ifip/item2012/
06.08.–10.08.2012 – Enschede/
The Netherlands
The 10th IFIP International Working
Conference on Information Technology in Educational Management
09.09.–12.09.2012 – Konstanz
Mensch und Computer 2012
MC 2012
http://hci.uni-konstanz.de/mc2012/
12.09.–14.09.2012 – Konstanz
Sportinformatik 2012
http://www.sportinformatik2012.de
16.09.–21.09.2012 – Braunschweig
Was bewegt uns in der/die Zukunft?
Gemeinsame Jahrestagung von GI
und GMDS
INFORMATIK 2012
www.informatik2012.de
17.09.2012 – Düsseldorf
CamelCaseCon: groovy
http://www.camelcasecon.de
19.09.–21.09.2012 – Ghent/Belgium
International Conference on Research and Practical Issues of
Enterprise Information Systems
CONFENIS 2012
http://www.confenis2012.be
24.09.–26.09.2012 – Hagen
Die 10. e-Learning Fachtagung
Informatik
DeLFI 2012
http://delfi2012.de
24.09.–27.09.2012 – Saarbrücken
KI Jahrestagung 2012
www.dfki.de/KI2012
26.09.–29.09.2012 – Bremen
11th International Conference on
Entertainment Computing
ICEC 2012
http://icec2012.org
12.11.–14.11.2012 – Magdeburg
Vision, Modeling and Visualization
Workshop 2012
VMV 2012
http://vc.cs.ovgu.de/vmv
14.11.–15.11.2012 – Karlsruhe
Automotive – Safety&Security 2012
http://www.ada-deutschland.de
14.11.–16.11.2012 – Bielefeld
Software Management 2012
http://www.swm2012.de
08.09.–11.09.2013 – Bremen
Mensch und Computer 2013
MC 2013
[email protected]
16.09.–20.09.2013 – Koblenz
INFORMATIK 2013
[email protected]
25.09.–28.09.2012 – Magdeburg
31st International Conference on
Computer Safety, Reliability and
Security
SAFECOMP 2012
http://www.ovgu.de/safecomp
Informatik_Spektrum_35_2_2012
167
HAUPTBEITRAG / MASSIVELY DISTRIBUTED SYSTEMS
}
Organic Design of Massively
Distributed Systems: A Complex
Networks Perspective
Ingo Scholtes · Claudio Juan Tessone
Introduction
Networked computing systems are becoming increasingly large, complex and – at the same time
– important for our everyday lives. Many of the
services we rely on, are now being collaboratively
provided by thousands or millions of machines
in large Peer-to-Peer (P2P) systems or data centers. Sustaining the robustness and manageability
of such systems are challenging tasks. Because of
the ongoing miniaturization of network devices,
their price-decline as well as the proliferation of
mobile and embedded computing equipment, scenarios in which billions of devices are connected
to global-scale information systems become reality.
Promising aspects of the coalescence of the virtual
and physical world that results from the increasing
incorporation of communication technology into
everyday objects, as well as the associated technical
and societal challenges have been highlighted in the
visions of Ubiquitous Computing [46] or the Internet
of Things [30].
Building services and applications in an environment of numerous dynamic and error-prone
communication devices poses enormous technical
challenges in terms of scalability, efficiency, manageability, and robustness. It is frequently argued
that, in order to cope with these challenges, computing technologies need to adopt the remarkable
self-organization, self-adaptation and self-healing
qualities of biological systems. Facilitated by
advances in the study of principles underlying selforganization mechanisms as well as the massively
increasing complexity of technical infrastructure,
in recent years the vision of Organic Computing has
been gaining momentum. The development of as-
sociated technologies is likely to go hand-in-hand
with a cutback of sophisticated algorithmic schemes
and deterministically organized communication
structures. Instead, adopting a heterodox approach
that utilizes simple mechanisms that underlie selforganization, adaptivity, robustness, and resilience
in natural systems is reasonable.
Driven mainly by the availability of massive
data sets, during the last decade these principles
have been studied in a variety of different contexts, including disparate fields like biology, physics,
computer science, economics, and sociology. The
resulting interdisciplinary strand of research is
subsumed as complex systems science and here we
argue that it offers a promising and quickly evolving methodological framework for the modeling,
design, and control of organic computing systems.
Providing a set of tools and abstractions to analyze
the collective properties of systems comprised of
a large number of stochastic, interacting elements,
complex systems science addresses one of the key
questions emerging in the scenario described above:
How can we analyze, monitor, and control the structure and dynamics of massively distributed systems
evolving from distributed mechanisms?
In this article we address some aspects that arise
in the management of robust and adaptive overlay
topologies for massively distributed systems. Here
we adopt the perspective of statistical mechanics
DOI 10.1007/s00287-012-0597-4
© Springer-Verlag 2012
Ingo Scholtes · Claudio Juan Tessone
ETH Zürich, Chair of Systems Design,
CH-8032 Zürich, Switzerland
E-Mail: {ischoltes, tessonec}@ethz.ch
Informatik_Spektrum_35_2_2012
75
{ MASSIVELY DISTRIBUTED SYSTEMS
Abstract
The vision of Organic Computing addresses
challenges that arise in the design of future
information systems that are comprised of numerous, heterogeneous, resource-constrained
and error-prone components. The notion organic highlights the idea that, in order to be
manageable, such systems should exhibit selforganization, self-adaptation and self-healing
characteristics similar to those of biological
systems. In recent years, the principles underlying these characteristics are increasingly being
investigated from the perspective of complex
systems science, particularly using the conceptual framework of statistical physics and
statistical mechanics. In this article, we review
some of the interesting relations between statistical physics and networked systems and discuss
applications in the engineering of organic overlay networks with predictable macroscopic
properties.
on the emergence of complex structures and collective dynamics in networks – an area that has
been particularly active and successful during the
last decade. In Sect. “Overlays, Random Graphs,
and Complex Networks”, we summarize structured
and unstructured approaches to the management of
overlay networks. Here we additionally review the
relevance of random graph theory for the design
of unstructured systems and introduce the relations between the study of statistical mechanics
and complex networks as well as the modeling of
dynamical processes. In Sect. “Managing Organic
Overlays – A Thermodynamic Perspective”, we discuss abstractions from statistical mechanics and
statistical physics in the design of organic overlay
networks. In Sect. “Conclusion and Outlook”, we
summarize challenges and opportunities of using
complex systems science in the engineering of distributed systems with predictable and controllable
self-* properties.
Overlays, Random Graphs,
and Complex Networks
Overlay networks – which define virtual connections
on top of physical communication infrastructures
– are becoming an increasingly important issue. As
76
Informatik_Spektrum_35_2_2012
argued in [45], the possibility to define communication topologies and protocols at the application
layer without having to make a – potentially globally
– coordinated change of existing protocols, standards, and communication infrastructures is an
important factor for a quick proliferation of novel
services on the Internet as well as in large-scale
data centers. The research of overlay topologies as
well as efficient distributed algorithms providing
core functionality like search, routing, and content dissemination has received a lot of attention
recently.
Most of this research has been done in the context of P2P systems, which are now increasingly used
for the cost-efficient distribution of data for example
by means of the BitTorrent protocol, the provision of
video-telephony services like Skype, or even to face
challenges emerging in large-scale scientific setups
like the Large Hadron Collider[43]. One usually distinguishes structured and unstructured approaches
in the management of overlay topologies. Most of
the currently deployed systems belong to the former
category. In such structured systems, virtual connections between machines are created in a globally
consistent way to construct a particular network
topology. While this allows for the development of
highly efficient algorithms for distributed search,
routing, or information dissemination, the major
difficulty is to maintain this fine-tuned topology
under dynamic conditions. Reconsidering the scenario outlined in Sect. “Introduction”, maintaining
fine-tuned structures will entail massive complexities due to the excessive fluctuation of participating
devices and the associated concurrency. In fact, for
the distributed hash table Chord it has been argued
in [7] that in settings with very large numbers of
highly dynamic participants, the communication
overhead imposed by mere topology maintenance
and management schemes could exceed the cost
for actual data transfer operations and thus dominates performance. It has further been argued that
designing, implementing, and debugging topology
maintenance schemes pose a huge challenge due
to the massive concurrency that is introduced by
failing or joining machines. These problems of structured overlays are well known in the literature and
question their usability in future scenarios like the
one laid out in Sect. “Introduction”. Hence, alternative approaches for dealing with large and dynamic
settings are being studied.
Unstructured Topologies
and Random Graph Theory
A straight-forward idea is to use unstructured overlays in which virtual connections between machines
are created in a simple, uncoordinated fashion
while still allowing all machines to communicate
with each other. While this reduces the overhead of
topology management, it necessitates probabilistic algorithms for example for distributed search
or routing that make no – or at least less specific
– assumptions about the structure of the network
or the placement of data items. Such schemes are
inevitably less efficient compared to those tailored
for a particular network structure. Nevertheless,
they are significantly simpler to implement and
allow for larger degrees of freedom in terms of
adapting the network structure to operational
conditions.
In terms of modeling performance and robustness, most unstructured approaches to the
management of overlays rely – either explicitly or
implicitly – on results from the field of random graph
theory which was established more than 50 years
ago [20]. In order to explain the analogies between
large, dynamic networked systems and statistical
mechanics, we briefly recall one of the basic models
of random graph theory. The so-called G(n, p) model
defines a probability space that contains all possible
graphs or networks1 G with n nodes. Assuming that
edges between pairs of nodes are being generated by
a stochastic process with uniform probability p, the
G(n, p) model assigns each network G with n nodes
and m edges the same probability to be created:
n(n–1)/2–m
PG (n, p) = pm · 1 – p
This simple stochastic model for networks has been
used in the modeling of a variety of real-world
networks. In particular, one can use it to make
predictions about the properties of unstructured
overlays, if virtual connections are assumed to be
created at random with probability p or, alternatively,
if an average number of p · n(n – 1)/2 connections
are established between randomly chosen pairs of
nodes.
In general, in the study of random networks one
is particularly interested in properties that hold for
a subset of network realizations whose probability
1
Throughout this article, we will use the terms graph and network
interchangeably.
measure converges to 1 as the size of the generated networks (in terms of the number of nodes)
increases. In this case one can say that a property
holds asymptotically almost surely for a randomly
generated network. This is because the probability
to draw a network that does not exhibit the property in question quickly vanishes. An authoritative
overview of the interesting results derived from this
perspective can be found in [14]. Two well-known
examples of particular relevance for the design of
overlay networks are results on the critical percolation threshold and the diameter. The critical
percolation threshold refers to a point in the G(n, p)
model’s parameter p above which the generated networks almost surely contain a connected component
that is of the order of the network size. For the G(n, p)
model it has been found that connected components
of a random graph are with high probability of the
order log(n) if p < 1/n. For p > 1/n the connected
component is of the order n [20]2 . In practical terms,
this result is a crucial prerequisite for the feasibility of unstructured overlay management schemes
since it tells that – if at least a certain minimum
number of connections is created in a random and
uncoordinated fashion – all machines will be able
to communicate with each other with a high probability. Another set of results that are important for
overlays with random structures relates the parameter p to the diameter of the resulting topology. It
further gives a criterion for the emergence of socalled small-world topologies which are assumed
to have a diameter of the order of the logarithm of
the network size. For the G(n, p) model, it has been
shown that the diameter is with high probability of
order log(n)/log(np), if the average number of links
per node is at least 1. In the design of unstructured
topologies, this argument is crucial to reason about
the efficiency of search and routing schemes.
Statistical Mechanics of Complex Networks
As argued in [15], the existence of so-called critical
points in the G(n, p) model’s parameter p and the
associated sudden change of macroscopic qualities
like diameter or connectedness, highlights interesting relations to phase transition phenomena in
statistical physics, i. e., sudden changes of material
properties as aggregate control parameters (e. g.,
2
Interestingly this is a so-called double-jump transition,
i. e., for p = 1/n the
2
size of the connected component is of the order n 3 .
Informatik_Spektrum_35_2_2012
77
{ MASSIVELY DISTRIBUTED SYSTEMS
temperature or pressure) change slightly. In recent years, these analogies to fundamental natural
phenomena have been substantially deepened by
reframing the study of random graph structures in
terms of statistical mechanics and statistical physics
(see e. g., [2, 11, 19, 21, 23, 37]). This perspective is
possible since statistical mechanics reasons about
configurations of many-particle systems, just like
random graph theory reasons about network configurations. Each of these particle configurations –
the so-called microstate – fixes the exact positions
and energy states of all particles present in a given
volume of space at a given temperature and total
energy. On the basis of energy distributions, particle
positions and fluctuations induced by temperature
as well as a quantum approximation, each microstate
can be assigned a probability based on combinatorial arguments. The set of all possible microstates
thus defines a probability space which is called a
statistical ensemble. The study of material properties
translates to reasoning about the probability measures for a subset of these particle configurations,
just like random graph theory reasons – for instance
– about the probability of the subset of network
configurations with a given diameter.
On the basis of these similarities, it has been
shown for instance in [23] that the G(n, p) model of
classical random graphs can be reframed in terms
of the so-called grand-canonical ensemble of manyparticle systems. In this perspective, the study of
statistical ensembles with fixed thermodynamic
quantities like volume, chemical potential, and temperature in statistical mechanics translates to the
study of network ensembles with fixed aggregate
statistics like a given number of nodes or edges, degree distributions, degree-degree correlations, or
clustering coefficients. In the resulting ensembles all
realizations with the same aggregate statistics (e. g.,
all networks with a particular degree sequence) are
assumed to have equal probability. In statistical mechanics, this corresponds to an adiabatic situation
in thermodynamic equilibrium while at the same
time the accessible states are being constrained by
certain fixed quantities. In the remainder of this
article we will thus refer to such probability spaces
as constrained adiabatic ensembles.
During the last decade such constrained adiabatic ensembles of networks have been studied
extensively in the fields of complex networks and
statistical mechanics [2, 34]. A particularly active
78
Informatik_Spektrum_35_2_2012
strand of research in this direction is the study of ensembles with fixed degree distributions following,
for instance, a power law. This is, the probability that a randomly chosen node in the network
has exactly k links, is proportional to k–γ for some
γ ∈ [2, ∞). Since the classical G(n, p) model can be
viewed alternatively as an adiabatic ensemble with
a fixed Poissonian degree distribution, this naturally extends earlier works on random graphs. Since
power-law degree distributions have been observed
for a number of real-world networks [2], the associated constrained adiabatic ensemble effectively
serves as a null model for these kinds of systems.
Over recent years, a rich set of results both on
the collective properties of such networks, as well as
on simple local mechanisms by which they emerge
have been obtained. Extensive surveys of results
based on this statistical physics perspective on complex network structures can be found in [10, 12].
Prominent results for the special case of networks
with heavy-tailed degree distributions include their
resilience against random faults [16] and targeted
attacks [17] or the performance of probabilistic distributed search schemes [1]. In the remainder of this
article, we illustrate that these results and – more
importantly – the underlying methodological framework are relevant for the design of robust networked
systems with organic properties.
Dynamical Processes in Complex Networks
So far, we have commented on the structural properties emerging in networks being drawn from
constrained adiabatic ensembles. For the design of
distributed algorithms which must operate in an
efficient and reliable way in large dynamic overlays,
it is equally important to have tools at hand that
allow one to reason about dynamical processes (like
e. g., information dissemination, synchronization,
or distributed search) operating upon them. In order
to formalize the problem, a useful representation
of a network is its adjacency matrix A where each
element is aij = 1 (aij = 0) if the nodes i and j are
connected (respectively, disconnected). Then, the
spectrum of such a network is given by the set of n
eigenvalues of its adjacency matrix. For the G(n, p)
model, it is possible to characterize the spectrum
of the networks in the limit of diverging network
sizes. In this model, and if there is a giant cluster that
spans the complete network, the probability p(λ) of
finding an eigenvalue λa in the spectrum follows the
so-called semi-circle law [31]:
⎧√
⎨ 4 n p(1–p)–(λa )2 if |λa | < 2n p(1 – p)
2π n p(1–p)
p(λa )=
⎩0
if |λa | ≥ 2 n p(1 – p)
The bulk of the distribution of eigenvalues is
centered around the null eigenvalue, with a charac√
teristic size proportional to n. However, the largest
eigenvalue λa1 , is proportional to p · n.
In order to study continuous dynamical processes in networks, we consider that each node i is
endowed with a continuous variable xi which describes its current dynamical state. Then, the change
of its state can be thought to be given by
n
d
xi = fi (xi ) + C
ij · h(xj ),
dt
j=1
(1)
where f (xi ) is a function describing a deterministic change of state of node i given its current state,
and h(xj ) is a coupling function given the state of
node j, and C is the coupling strength. In (1), ij are
the elements of the so-called Laplacian matrix, L.
Such a matrix is defined as ij = – ki δij + aij , where
ki is the degree of node i, and δij are the elements
of the identity matrix. The Laplacian matrix naturally extends the Laplacian operator ∇ 2 – as used in
the description of dynamics in spatially extended,
physical systems – into a discrete manifold.
Different kinds of collective behavior have been
observed in the study of dynamical processes in
networks emerging from different constrained adiabatic ensembles [5, 12]. Whenever different nodes
show the same dynamics, it can be said that a synchronized state has emerged. In general, it was
shown [9] that such synchronized states are stable if λln /λl2 < β, where λln and λl2 are (respectively)
the largest and smallest nonzero eigenvalues of the
Laplacian matrix, and β is given by the functions
f (x) and h(x) describing the dynamical properties
of the system. This means that there is a relationship between structural properties of the network
(in terms of its eigenvalues) and dynamical process
taking place at the level of nodes as to whether the
system is able to synchronize. This runs against typical intuition on synchronization phenomena, stating
that if the coupling strength is large enough, the
system should exhibit a synchronized state. Furthermore, if a synchronized state emerges the question
of how much time the system needs to reach such
a state emerges. For arbitrary network ensembles
it was shown [4] that the consensus time is of the
order TC = (ln C – ln /2)/λl2 , where C is an integration constant (related to the initial conditions) and
is a synchronization threshold, i. e., the difference
below which nodes are assumed to be synchronized.
Statements on the impact of the Laplacian
spectrum play an important role when assessing
the properties of synchronization and consensus
schemes applied in distributed systems and relating
the emerging collective dynamics to the topology of
the network. In structured and unstructured overlays, synchronization models have been proposed
to provide a network-wide synchronous notion of
time epochs or protocol cycles [6, 8, 29, 42]. In fact,
it is even possible to infer from the synchronization
dynamics at the level of individual nodes statements
about the cluster structure and the Laplacian matrix
of the network [41]. Finally, regarding detrimental
collective phenomena like the synchronization of periodic routing messages described in [22], an analysis
of the spectral properties of networks is crucial.
Dynamical processes additionally can take place
in a discrete space. Perhaps the simplest of such processes are random walks in a network. The question
how fast a node receives and spreads information in
such a random process is important for example in
the study of stochastic search and transport phenomena[35]. To quantify this, the random-walk
centrality Ci was introduced which – for node i –
is given by:
–1
ki
ki
Pii (t) – Ci = i ki
i ki
Here, Pii (t) is the probability that a random walk
which started at i at time zero, returns to the same
node after t time steps, and can be readily computed
by means of the adjacency matrix A. This index
shows that nodes with large centrality receive information faster. In the case of limited bandwidth
capacities, it indicates that such nodes may become
overloaded first.
Arguments about the relation between spectral
properties of networks and the dynamics of random walk-related processes have been used – again
implicitly and explicitly – in a variety of contexts
like multicast communication [24], database replica
maintenance [18], the computation of networkwide aggregates [26], or random sampling [48]. By
means of the spectral perspective on the equivalent
Informatik_Spektrum_35_2_2012
79
{ MASSIVELY DISTRIBUTED SYSTEMS
of the Laplacian operator in networks, it is possible to study the performance of these schemes for
unstructured overlays given that they are drawn
from a particular statistical ensemble. This clearly
demonstrates the relevance of these techniques for
the design of organic networked systems, while at
the same time highlighting interesting mathematical
relations to fundamental natural phenomena being
studied in statistical physics.
Managing Organic Overlays –
A Thermodynamic Perspective
From an engineer’s perspective, being able to assess
the structural and dynamical properties of networks
with complex structures sounds appealing. However, since most of the findings are based on simple
stochastic models for complex systems one needs to
be careful when applying them to real-world systems: Statistical ensembles of complex networks
should be viewed as mere null models for networked
systems with complex structures and it is unlikely
that they accurately reproduce the properties of sophisticated infrastructures like the Internet, which
are subject to numerous technological constraints.
Some of the fallacies that can arise when imprudently
applying oversimplified complex network models to
sophisticated technical infrastructures have been
summarized for instance in [47].
Nevertheless, given that the statistical mechanics perspective on networks is able to reason about
structural and dynamical properties that are relevant
for the design and operation of overlay networks, it
is reasonable to study how one can use models for
complex networks in a constructive rather than in an
explanatory way. Overlay management schemes can
explicitly be designed based on a stochastic model
that gives rise to a class of topologies whose properties are advantageous for a particular setting. By
means of distributed probabilistic protocols – like
for instance suitably configured random walk sampling or rewiring schemes – reproducing a stochastic
model is often much simpler than implementing and
debugging complex algorithms that precisely control
the topology.
The perspective of statistical mechanics actually
allows one to contrast this approach with traditional
structured and unstructured overlays: From this
point of view structured approaches give rise to states
of small (statistical) entropy in the sense that only
a small subset of all possible network realizations
80
Informatik_Spektrum_35_2_2012
are accessible. This maximizes the amount of information one has about the detailed structures of the
network. This information can then be used to design algorithms that utilize the network structure
to provide efficient key lookups, routing, and information spreading. However, maintenance schemes
are required to prevent a gradual loss of structure –
and thus increase of entropy – due to the dynamics
of users and devices, hard- and software failures or
communication errors. As such, the overhead induced by topology maintenance mechanisms can
be viewed in analogy to the input of energy that is
used by nonequilibrium biological systems to prevent the entropy increase that is due to the second
law of thermodynamics. Analogously, unstructured
approaches can be related to states of maximum statistical entropy in which all network realizations
are equally likely, as the topology is constructed in
a completely uncoordinated fashion. In this case, distributed algorithms can not use a priori information
about the detailed structures of the network structure and therefore flooding or exhaustive search are
the only viable options.
One of the most interesting aspects of complex
network science is that it allows one to explore
the interesting middle-ground of statistically or
thermodynamically structured topologies with intermediate levels of entropy. These networks are neither
completely random nor deterministic. They rather
introduce a statistical structure (like a particular
type of degree sequence, a certain clustering coefficient, or correlations between data location and
network structures) that facilitates adaptivity and
that allows one to solve algorithmic tasks more efficiently than in unstructured systems. For distributed
search in P2P systems, it has been shown in [39]
that generic structures of randomly generated networks with a power law degree distribution can be
exploited in order to improve search performance,
while being oblivious to the details of the topology. Similar results have been obtained for routing
schemes making use of correlations between node
addresses and network structure, particular clustering structures, or an embedding into Euclidean or
Hyperbolic coordinate spaces [13, 27, 36, 38, 44].
Enforcing Ensembles
In Sect. “Statistical Mechanics of Complex Networks”, we commented on the rich body of results on
collective properties like connectedness, diameter,
resilience against faults and attacks, as well as on
the performance of dynamical processes like information dissemination, synchronization, consensus
and gossiping schemes. Since these statements are
derived based on statistical ensembles, they are
necessarily stochastic. However, statements on properties that hold asymptotically almost surely become
more reliable as the size of the network topology
(in terms of the number of nodes) increases, just
like statements on properties of thermodynamic
systems become more reliable as the volume being
considered is increased. This applies for instance to
stochastic guarantees on the diameter or connectedness of overlay networks emerging from suitable
stochastic processes. Again, this is due to the fact
that it is becoming increasingly unlikely to generate
a network realization from a set whose probability
measure converges to zero under the given construction process3 . This often contrasts with the kind of
guarantees one can obtain for structured overlay
networks.
So far, we have discussed the link between the
abstraction of constrained adiabatic ensembles of
networks and collective network properties. However, from a practical perspective this link is useless
if we cannot relate the ensemble description of
a system to the distributed processes shaping the
overlay at the level of individual nodes. In real-world
systems, one may be confronted with situations in
which the actual stochastic dynamics constructing
the overlay depend on external factors that cannot easily be influenced. Here one can often use
tools from statistical mechanics (like e. g., master
equations or mean-field approximations) to derive aggregate statistical quantities of interest –
and thus the ensemble description of the system –
from a stochastic description of individual nodes
(see e. g., the application of this procedure in [3]).
The situation is different when we wish to actually
design a topology management scheme that constructs a network topology drawn from a particular
constrained adiabatic ensemble which has desirable
properties. In this case, one can often employ a con-
figuration approach to analytically derive the local
stochastic dynamics from the ensemble description.
In the following we will briefly demonstrate this
approach for the particular case of random scalefree networks. The actual distributed mechanism
has been presented and evaluated in much detail
in [40, 41].
The basic idea is to start with an arbitrary connected topology that has been generated for instance
by a bootstrapping process. One can then progressively rewire all connections by means of a biased
random walk sampling scheme like the one proposed in [48]. Again referring to [40, 41] for a more
detailed algorithmic description of the protocol,
each rewiring of an existing connection e between
two nodes i and j in the topology is initiated by one of
their endpoints. This node deletes the connection e
and creates a sampling message that contains the addresses of i and j. By means of two consecutive biased
random walks of length l, two new nodes v and w (the
final nodes at which the two random walks reside
after l steps) are sampled. In a distributed setting,
the address of the first node v sampled by the first
random walk can be incorporated in the sampling
message passed along in the second random walk.
In this case, the final node w has all information
needed to establish a new overlay link e = (v, w) if
it does not exist already. The process is illustrated in
Fig. 1, which shows a random walk rewiring of edge
e initiated by node 0. A first random walk takes three
steps and finds node v = 3. After another 3 steps
a second endpoint w = 6 is found and the new edge
e = (3, 6) is created. From a statistical mechanics
perspective, this scheme results in the fact that edges
3
Since in practice one necessarily deals with finite-size systems, it is important
to note that technically one also needs to consider how fast the associated
probability converges as the network size increases. It is often possible to give
analytical expressions for these so-called finite-size effects. While we refer the
interested reader to [19] for more details, at this point it is sufficient to note
that the properties mentioned above hold for networks whose size is reasonably
small (of the order of a few hundred to a few thousand nodes) to be of practical
value for the envisioned scenario.
Fig. 1 Rewiring of edge e initiated by node 0 by means of two
consecutive random walks with step length three. The red
dotted edge is replaced by the green dotted one
Informatik_Spektrum_35_2_2012
81
{ MASSIVELY DISTRIBUTED SYSTEMS
mimic a random particle motion according to an engineered energy landscape which in this particular
case is defined according to the adiabatic ensemble
of scale-free networks.
If the transition kernel of the random walk sampling is configured appropriately and the random
walk is sufficiently long, the topology is guaranteed
to be drawn from the desired adiabatic ensemble
once all connections have been rewired. Here, the
number of random walk steps needed to allow the
system to equilibrate (i. e., forget about the arbitrary
initial topology) is crucial to apply equilibrium arguments about the targeted ensemble. In experiments
described in [40], we have found a length of log(n)
steps to be sufficient where n is the number of nodes.
An analytical approach that is based on the framework described in Sect. “Dynamical Processes in
Complex Networks” can be used to derive an upper
bound for the required random walk length (see details in [40]). On the basis of a configuration model
approach for a constrained adiabatic ensemble of
scale-free networks that was introduced in [28] as
well as the Metropolis–Hastings algorithm [25] we
can actually derive the required transition kernel
for the random walk. In order to effectuate an ensemble with a particular exponent γ one has to
configure the random walk such that each node i
forwards a sampling message to a neighbor j with
probability
Pi,j =
di
dj
γ 1–1
i
j
(2)
where i, j are numeric, not necessarily unique node
identifiers chosen uniformly at random from an identifier space. The effect of the resulting rewiring process on the degree distribution is shown in Fig. 2
in terms of the evolution of the fitted degree distribution exponent as well as the Kolmogorov–Smirnov
statistic D which quantifies the goodness of the fit.
The decreasing values for D show that the hypothesis that the topology is indeed a random network
with a power-law degree distribution becomes more
reasonable. Since the network is sampled at random,
for this simple topology management mechanism
one can thus safely rely on all analytical results
that hold for an adiabatic ensemble of random
scale-free networks with the chosen exponent.
The Micro-Macro Link
An interesting aspect of the procedures described
above is that they allow one to analytically link
a stochastic model for the micro-scale dynamics
of an overlay (at the level of individual nodes and
protocol messages) with the collective properties
that are observable at the system level. The three
levels that naturally arise in this perspective are
depicted in Fig. 3. At the microscopic level, one assumes a stochastic model that captures the behavior
of individual nodes with respect to the way connections in the overlay are being constructed. In the
previous section, we discussed that such a stochastic
micro-model can be related to an adiabatic statistical ensemble – which defines the next level of
description – and vice-versa. At the level of statistical
ensembles, aggregate network statistics like average
Fig. 2 Time evolution of
5000 node networks during
adaptation runs with
γ ∈ [2.1, 3.5]
82
Informatik_Spektrum_35_2_2012
Fig. 3 Micro-macro link provided by a thermodynamic approach
to the design of overlay networks
degree, system size, degree distributions or other
correlations effectively serve as thermodynamic
quantities that constrain the accessible states and
thus determine the structural and dynamical properties of the system. By measuring the probability
of network configurations with particular macroscopic characteristics, one can then make strong
probabilistic statements about collective properties
like diameter, connectedness, or spectral properties
which emerge at the macro-level of the system.
At the intermediate level of statistical ensembles,
rather than a single ensemble, one may further consider a set of ensembles with different constraints.
By means of the micro-macro link exemplified in the
previous section, it is possible to provide distributed
adaptation strategies which switch between overlay
topologies drawn from different ensembles based
on the environmental conditions. Reconsidering the
particular distributed sampling scheme exemplified
above, the exponent parameter γ actually defines
a point in a continuum of adiabatic ensembles of
scale-free networks that can be tuned by changing
the parameter in the random walk bias, thus effectively changing the stochastic connection rewiring
dynamics.
Adaptation in Organic Overlays –
Triggering Phase Transitions
While the possibility to switch between different
ensembles by tuning a sampling bias can be useful
per se (e. g., to switch between random and scale-free
topologies), particularly interesting aspects that can
be utilized are phase transition phenomena, i. e., the
existence of critical points in the space of control
parameters at which the macroscopic properties of
the resulting topology change suddenly. This may involve smooth (second-order) or abrupt (first-order)
changes as well as phase transitions with hysteresis
effects. Such critical points are of primary interest
for the complex networks community and a number
of analytical results have been obtained that relate
these points to a qualitative change of macroscopic
network properties. For scale-free networks, such
phenomena have been studied in detail. When considering a continuum approximation of scale-free
networks (justified in the thermodynamic limit of
infinite systems), the probability for a node to have
exactly k connections is given by a Zeta (also called
scale-free) distribution,
P(k) =
k–γ
ζ(γ )
with ζ : R → R being the real-valued Riemann
–γ
Zeta function, ζ(γ ) = ∞
i=1 i . Many results for
critical phenomena in networks with fixed degree
distributions are due to the so-called Molloy–Reed
criterion [32] which links the relation of the distribution’s first two moments to the existence of a giant
connected component. This has been used successfully to study – as mentioned before – the error and
attack tolerance of random power-law networks. For
this special case, in [16] it was found that at least
a fraction
–1
ζ(γ – 2)
r := 1 –
–1
ζ(γ – 1)
of nodes need to fail at random for the giant connected component of a power-law network to be
destroyed. Regarding the control parameter γ –
which determines the behavior of the distributed
rewiring scheme described in Sect. “Enforcing Ensembles” – the convergence behavior of the Zeta
function results in a phase transition once the parameter crosses the critical point γ = 3. For γ ≥ 3,
the terms ζ(γ – 2) and ζ(γ – 1) are constants, thus
resulting in a constant nonzero value for the critical fraction r. For γ ∈ (2, 3), the term ζ(γ – 2)
diverges. In this range r → 1, i. e., almost all nodes
can be removed at random without destroying the
giant connected component. The parameter γ thus
effectively allows one to produce a continuum of
constrained adiabatic ensembles while at the critical
point γ = 3, the resilience properties of the resulting
networks undergo a qualitative change.
Informatik_Spektrum_35_2_2012
83
{ MASSIVELY DISTRIBUTED SYSTEMS
The same argument about the change in the convergence behavior of the Zeta function has actually
been used to derive phase transitions in terms of
diameter, attack resilience [17] or the efficiency of
spreading phenomena [33]. Given that – depending
on the used distributed algorithms, as well as the
current operational conditions – these properties
can be both desirable or detrimental. So, the knowledge about these effects can actually be used for an
active adaptation of collective network qualities. For
the random walk rewiring described in Sect. “Enforcing Ensembles”, all one has to do is to change the
control parameter γ in (2). Figure 4 shows simulation results that have been obtained based on this
idea and that were originally presented in [40]. Here,
the random rewiring protocol described above has
been applied continuously, while at certain times
(indicated by vertical lines in Fig. 4a and b) the parameter γ was changed. After each modification, the
network was allowed to equilibrate by progressively
resampling all connections according to the new
constrained adiabatic ensemble. Figure 4a shows the
evolution of the fitted degree distribution exponent
γ . The Kolmogorov–Smirnov statistic D depicted
in Fig. 4b shows the goodness of the assumption
that the connectivity distribution indeed follows
a power law with the fitted exponent. Smaller values
for D again represent a larger reliability of the fit.
Here, it can clearly be seen that near the end of the
adaptation cycles (which was at the same time the
beginning of a new one with a different γ ), the degree distribution of the network does indeed follow
a power law.
On the basis of the theoretical findings regarding the critical point of γ = 3 for the robustness
of the topology against attacks, one would expect
that the robustness changes qualitatively between
the last two adaptation rounds, i. e., when the ex-
Fig. 4 Time Evolution of
scale-free network during
multiple adaptation cycles.
Start/end times of
adaptation cycles are
indicated by vertical lines
84
Informatik_Spektrum_35_2_2012
ponent of the topology is increased from 2. 1 to 3.5.
To exemplify this theoretical result, Fig. 4c,d shows
the network topology that remains after 10% of the
most connected nodes have been removed from a
300 node network at these two different points in
the adaptation process. The large number of isolated
nodes and clusters in Fig. 4c and the comparison
with the topology shown in Fig. 4d clearly show
the practical implications of this theoretical finding
in terms of resilience. In summary, this particular
phase transition effect in equilibrium ensembles of
scale-free networks can be used to make a tradeoff between desirable and detrimental properties.
In phases where efficient spreading is needed, the
topology can be sampled from an adiabatic ensemble with γ ∈ (2, 3). Similarly, according to [39]
γ ∈ (2, 2.3) should be chosen to maximize the efficiency of a random walk-based distributed search
in scale-free topologies. In situations where attacks
or a spreading of failures are being detected, the
local connection sampling can be instrumented such
that the resulting topology is much more resilient
against these effects, while at the same time reducing
the efficiency of distributed search and information
dissemination.
Conclusion and Outlook
In this article, we have summarized the statistical
mechanics and the statistical physics perspective on
the modeling of complex network structures. We
then outlined some ideas on how this perspective
can constructively be used in the management of
overlay networks for very large, dynamic systems.
In particular, we argue that – at least in very large
and highly dynamic systems – it can be easier to
enforce a particular statistical ensemble which will
give rise to desirable macro-level properties and
performance of distributed schemes than using
sophisticated topology maintenance schemes. By
means of active randomization of protocols, one can
thus obtain strong, thermodynamic guarantees for
the structure and dynamics emerging in sufficiently
large systems. For the design of robust and adaptive
organic computing systems, we thus argue that randomization and loosening precise control are crucial
ingredients.
While first examples of actual distributed mechanisms that allow one to actively use some of the intriguing results of complex network science have
been given, this work is necessarily incomplete. For
the future, we envision for instance mechanisms that
make use of the natural dynamics of Peer-to-Peer
systems for the efficient construction and adaptation of an overlay topology with complex, yet
predictable structures and properties. For this, rather
than actively rewiring connections, one can use the
natural turnover of machines and users (an effect
usually called churn) and apply random connection
sampling schemes only as nodes join the system,
machines change their characteristics, or existing
connections fail. By assuming that the system is in
a state of equilibrium, the collective properties of the
overlay network can then be predicted in analogy to
the analysis of many-particle systems by means of
statistical mechanics. Moreover, within this framework it may be possible to reduce the costs induced
by structure maintenance protocols massively, thus
making such an approach suitable for large and
dynamic systems. Referring again to the scenario
depicted in Sect. “Introduction”, one may thus be
tempted to summarize the challenges of future systems, as well as the idea of addressing them in the
framework presented in this article in the following
way:
As network devices become more akin to particles
in terms of number, size, and stochastic behavior,
can we design distributed systems using models,
methods, and abstractions from statistical mechanics and thermodynamics?
Considering recent advances in the statistical
physics’ study of complex systems in general, and
complex networks in particular, one can argue that
this is true at least for some aspects of large dynamic
systems. An interesting prospect of this perspective
is the fact that – in contrast to particles in thermodynamic systems – we can actually program devices
to change their local dynamics in a meaningful way,
for instance to actively trigger transitions between
phases in which the network is more resilient or
allows for more efficient spreading or distributed
search.
Given the current surge of interest in the statistical mechanics approach to the modeling of
complex networks, we thus think that it is reasonable
to foresee a number of interesting applications in
the engineering of organic computing systems and
in the design of mechanisms for sustainable future
techno-social systems.
Informatik_Spektrum_35_2_2012
85
{ MASSIVELY DISTRIBUTED SYSTEMS
References
1. Adamic LA, Lukose RM, Puniyani AR (2001) Search in power-law networks. Phys
Rev E 64(4):046135
2. Albert R, Barabási A-L (2002) Statistical mechanics of complex networks. Rev Mod
Phys 74(1):47–97
3. Barabási A-L, Albert R, Jeong H (1999) Mean-field theory for scale-free random
networks. Physica A 272(1–2):173–187
4. Almendral JA, Díaz-Guilera A (2007) Dynamical and spectral properties of complex networks. New J Phys 9(6):187–187
5. Arenas A, Díaz-Guilera A, Kurths J, Moreno Y, Zhou C (2008) Synchronization in
complex networks. Phys Rep 469(3):93–153
6. Babaoglu Ö, Binci T, Jelasity M, Montresor A (2007) Firefly-inspired heartbeat synchronization in overlay networks. In: First International Conference on SelfAdaptive and Self-Organizing Systems, SASO’07, Boston, Mass., USA, July 9-11,
pp 77–86, IEEE, 2007
7. Balakrishnan H, Kaashoek MF, Karger D, Morris R, Stoica I (2003) Looking up data
in P2P systems. Commun ACM 46(2):43–48
8. Baldoni R, Corsaro A, Querzoni L, Scipioni S, Piergiovanni ST (2009) CouplingBased Internal Clock Synchronization for Large-Scale Dynamic Distributed Systems.
IEEE T Parall Distrib 99(RapidPosts):607–619
9. Barahona M, Pecora LM (2002) Synchronization in small-world systems. Phys Rev
Lett 89(5):54101
10. Barrat A, Barthélemy M, Vespignani A (2008) Dynamical processes on complex
networks. Cambridge University Press, New York, NY, USA
11. Berg J, Lässig M (2002) Correlated random networks. Phys Rev Lett 89(22):228701
12. Boccaletti S, Latora V, Moreno Y, Chavez M, Hwang DU (2006) Complex networks:
Structure and dynamics. Phys Rep 424(4-5):175–308
13. Boguñá M, Krioukov D, Claffy KC (2008) Navigability of complex networks. Nat
Phys 5(1):74–80
14. Bollobas B (2001) Random graphs. Cambridge Univ. Press, Cambridge
15. Cohen JE (1988) Threshold Phenomena in Random Structures. Discrete Appl
Math 19(1–3):113–128
16. Cohen R, Erez K, Ben-Avraham D, Havlin S (2000) Resilience of the internet to
random breakdowns. Phys Rev Lett 85(21):4626–4628
17. Cohen R, Erez K, Ben-Avraham D, Havlin S (2001) Breakdown of the Internet
under intentional attack. Phys Rev Lett 86:3682
18. Demers A, Greene D, Hauser C, Irish W (1987) Epidemic algorithms for replicated
database maintenance. In: Proceedings of the sixth annual ACM Symposium on
Principles of distributed computing, Vancouver, BC, Canada, August 10–12, 1987,
pp 1–12
19. Dorogovtsev SN, Ferreira Mendes JF, Samukhin AN (2003) Principles of statistical
mechanics of random networks. Nucl Phys B 666:396
20. Erdős P, Rényi A (1959) On random graphs I. Publ Math Debrecen 6(290–297):156
21. Farkas I, Derényi I, Palla G, Vicsek T (2004) Equilibrium Statistical Mechanics of
Network Structures. Lect Notes Phys 650:163–187
22. Floyd S, Jacobson V (1994) The Synchronization of Periodic Routing Messages.
IEEE/ACM T Netw 2:122–136
23. Park J, Newman MEJ (2004) Statistical mechanics of networks. Phys Rev E 70(6):
066117, http://pre.aps.org/abstract/PRE/v70/i6/e066117
24. Gupta I, Kermarrec AM, Ganesh AJ (2006) Efficient and adaptive epidemic-style
protocols for reliable and scalable multicast. IEEE T Parall Distrib 17(7):593–605
25. Hastings WK (1970) Monte Carlo sampling methods using Markov chains and
their applications. Biometrika 57(1):97
86
Informatik_Spektrum_35_2_2012
26. Jelasity M, Montresor A, Babaoglu Ö (2005) Gossip-based aggregation in large dynamic networks. ACM T Comput Syst 23(3):219–252
27. Kleinberg JM (2006) Complex Networks and Decentralized Search Algorithms. In:
International Congress of Mathematicians (ICM), 22–28 August 2006, Madrid,
Spain
28. Lee D-S, Goh K-I, Kahng B, Kim D-H (2004) Evolution of scale-free random graphs:
Potts model formulation. Nucl Phys B 696:351–380
29. Lucarelli D, Wang I-J (2004) Decentralized synchronization protocols with nearest neighbor communication. In: SenSys ’04: Proceedings of the 2nd international
conference on Embedded networked sensor systems, New York, NY, USA, ACM,
pp 62–68
30. Mattern F, Flörkemeier C (2010) Vom Internet der Computer zum Internet der
Dinge. Informatik-Spektrum 33(2):107–121
31. Mehta ML (1991) Random Matrices, 2 edn. Academic Press, New York
32. Molloy M, Reed B (1995) A critical point for random graphs with a given degree
sequence. Random Struct Algor 6(2–3):161–180
33. Moreno Y, Nekovee M, Vespignani A (2004) Efficiency and reliability of epidemic
data dissemination in complex networks. Phys Rev E 69(5):055101
34. Newman MEJ (2003) The structure and function of complex networks. SIAM Rev
45(2):167–256
35. Noh JD (2004) Random Walks on Complex Networks. Phys Rev Lett 92(11):4
36. Papadopoulos F, Krioukov D, Boguna M, Vahdat A (2010) Greedy Forwarding in
Dynamic Scale-Free Networks Embedded in Hyperbolic Metric Spaces. In: Proceedings of the IEEE Infocom Conference, March 19, 2010, San Diego, CA, USA
37. Reichardt J, Bornholdt S (2006) Statistical mechanics of community detection.
Phys Rev E 74(1):1–14
38. Sandberg O (2006) Distributed routing in small-world networks. In: Proceedings
of the eighth Workshop on Algorithm Engineering and Experiments and the third
Workshop on Analytic Algorithmics and Combinatorics, Society for Industrial
Mathematics, January 21, 2006, Miami, Florida, p 144
39. Sarshar N, Boykin PO, Roychowdhury VP (2004) Percolation search in power law
networks: Making unstructured peer-to-peer networks scalable. In: Proceedings
of the Fourth International Conference on Peer-to-Peer Computing, IEEE Computer Society, 25–27th August 2004, Zurich
40. Scholtes I (2010) Distributed Creation and Adaptation of Random Scale-Free Overlay Networks. In: Proceedings of the Fourth IEEE Conference on Self-Organizing
and Self-Adaptive Systems (SASO), September 27–October 1, 2010, Budapest,
Hungary, IEEE, pp 51–63
41. Scholtes I (2011) Harnessing Complex Structures and Collective Dynamics in Large
Networked Computing Systems. Dissertation, University of Trier
42. Scholtes I, Botev J, Esch M, Sturm P (2010) Epidemic Self-Synchronization in Complex Networks of Kuramoto Oscillators. Adv Complex Syst 13(1):33–58
43. Scholtes I, Kolos S, Zema PF (2008) The ATLAS Event Monitoring Service – Peerto-Peer Data Distribution in High-Energy Physics. IEEE T Nucl Sci 55(3):1610–1620
44. Thadakamalla HP, Albert R, Kumara SRT (2007) Search in spatial scale-free networks. New J Phys 9(6):190–190
45. Waldhorst OP, Bless R, Zitterbart M (2010) Overlay-Netze als Innovationsmotor im
Internet. Informatik-Spektrum 33(2):171–185
46. Weiser M (1991) The computer for the 21st century. Scientific American
47. Willinger W, Alderson DL, Doyle JC (2009) Mathematics and the Internet: A Source
of Enormous Confusion and Great Potential. Internet Res 56(5):586–599
48. Zhong M, Shen K, Seiferas J (2008) The Convergence-Guaranteed Random Walk
and Its Applications in Peer-to-Peer Networks. IEEE T Comput 57(5):619–633
{ AKTUELLES SCHLAGWORT* / MINING SOCIAL MEDIA
Mining Social Media
Martin Atzmueller
Social Media als aktuelles Phänomen
Eine ganze Reihe von Ereignissen wurden in
letzter Zeit zuerst getwittert, bevor sie über die
,,klassischen“ Medien verbreitet wurden, z. B. die
Flugzeugnotlandung von US Airways Flug 1549 im
Hudson River [1] oder die Wahl des Bundespräsidenten 2009 [2]. Social-Media-Dienste wachsen rasant:
Beispielsweise zeigte das soziale Netzwerk Google+
(Start: Juni 2011) ein exponentielles Nutzerwachstum: 10 Mio. Nutzer in 16 Tagen und 20 Mio. in 24
Tagen [3]; Facebook (Start 2004) hat im Januar 2012
ca. 800 Mio. Nutzer [4].
Was macht nun die Faszination von Social Media
aus? Social Media erfasst nutzergenerierte Inhalte in
,,klassischen“ sozialen Diensten wie z. B. Twitter,
aber beispielsweise auch in RFID-basierten Anwendungen [5, 14]. Damit beinhaltet Social Media
weitere Dimensionen, wie sie sich in Sensornetzwerken, mobilen Geräten und dem ubiquitären Web
eröffnen. Aus Anwendungssicht spielt die Analyse
von Communities eine große Rolle, um gemeinsame Interessen oder spezielle Kompetenzen zu
identifizieren. Diese können dann z. B. zur Personalisierung, für Werbung und für Empfehlungen genutzt
werden. Weiterhin stellt sich für Unternehmen oft
die Frage, wie produktbezogene Kundenmeinungen bewertet oder Experten identifiziert werden
können [14].
Wie können wir nun aus all diesen lokalen, verteilten, und unterschiedlichen Social-Media-Daten,
die durch menschliche Kommunikation und Interaktion erzeugt werden, interessante Informationen,
Muster und letztendlich Wissen extrahieren? Zunächst wollen wir im Folgenden wichtige Begriffe
klären, bevor wir konkrete Techniken und Anwen-
132
Informatik_Spektrum_35_2_2012
dungen betrachten, wobei wir hier insbesondere auf
Communities eingehen.
Begriffe
Soziale Medien (Social Media): Kaplan und Haenlein definieren Social Media als internetbasierte
Anwendungen, die auf Techniken des Web 2.0 aufbauen. Kern ist die Erstellung, Gestaltung und der
Austausch nutzergenerierter Inhalte (user generated content) [11]. Letztere schließen dabei alle
Anwendungsmöglichkeiten von Social Media ein,
hinsichtlich der Gestaltung durch Endnutzer sowie ihrer Verfügbarkeit. Intuitiv verstehen wir
unter Social Media daher internetbasierte Systeme
und Dienste im ubiquitären Web, die als Medium
soziale Daten menschlicher Interaktion und Kommunikation bereitstellen. Dies schließt auch Daten
von Sensornetzwerken oder mobilen Geräten ein,
solange die Daten von realen Nutzern erzeugt
werden.
Soziale Netzwerkanalyse (Social Network Analysis): Im Kontext von Social Media betrachten
wir soziale (Online-)Netzwerke, die ,,eine besondere Form von Gemeinschaft sind [...], bei denen
die Interaktion und Kommunikation der Akteure
DOI 10.1007/s00287-012-0601-z
© Springer-Verlag 2012
Martin Atzmueller
Knowledge And Data Engineering Group, Universität Kassel,
Wilhelmshöher Allee 73, 34121 Kassel
E-Mail: [email protected]
*Vorschläge an Prof. Dr. Frank Puppe
<[email protected]> oder
Prof. Dr. Dieter Steinbauer <[email protected]>
Alle „Aktuellen Schlagwörter“ seit 1988 finden Sie unter:
www.ai-wuerzburg.de/as
durch eine technische Plattform und die Infrastruktur des Internets unterstützt wird. Verbindendes
Element ist dabei ein gemeinsames Ziel, Interesse
oder Bedürfnis, ...“ [10, S. 263]. Hier zeigt sich
der Bezug zu Communities (Gemeinschaften), die
solche verbindenden Elemente manifestieren. Im
Kern betrachtet die soziale Netzwerkanalyse [19]
eine Menge von Knoten (Akteure) und ihre Verbindungen, die die Beziehungen zwischen ihnen
modellieren, z. B. Freundschaftsbeziehungen oder
die ,,Follower“-Beziehung in Twitter. Mithilfe der
sozialen Netzwerkanalyse lässt sich beispielsweise
in der Epidemiologie analysieren, wie menschliche
Kontaktmuster die Verbreitung von Krankheiten
(z. B. Grippe) fördern oder verhindern [18].
Communities: Communities stellen als
,,Gemeinschaften“ intuitiv besonders ,,dicht verbundene“ Gruppen von Knoten im sozialen Netzwerk
dar. Die Verbindungen können dabei durch unterschiedliche Interessen, Bedürfnisse, u. ä. motiviert
sein. Es existieren verschiedenste Definitionen von
Communities, wobei die obige den Kern der meisten
Definitionen umfasst. Grundsätzlich wird zur Auffindung und zur Bewertung einer Community meist
die Vernetzung (Dichte) innerhalb der Community
mit dem Rest-Netzwerk betrachtet [8].
Techniken
Zur Analyse von Social-Media-Daten können einerseits deren Inhalte genauer untersucht werden, um
etwa mit natürlichsprachlichen Methoden (Natural
Language Processing, NLP) linguistische Analysen durchzuführen. Andererseits kann der Fokus
auch auf der induzierten Netzstruktur der betrachteten Social–Media-Daten liegen, beispielsweise
in einem Freundschaftsnetzwerk. In einem solchen Fall können durch Methoden der sozialen
Netzwerkanalyse (SNA) Untersuchungen auf der
Gesamtebene des Netzwerks durchgeführt werden [19], um etwa die durchschnittliche Pfadlänge
zwischen Knoten oder den Durchmesser des Netzwerks zu bestimmen. Weiterhin können auf Knotenund Netzwerkebene leicht verschiedene Zentralitätsmaße im Netzwerk bestimmt werden, die auf
die Bedeutung oder das Prestige einzelner Knoten
hindeuten, um wichtige Knoten (Hubs) zu identifizieren oder Rollen der Knoten zu analysieren [6, 19].
Beispiele sind die Gradzentralität (degree centrality) als Anzahl der Verbindungen eines Knotens zu
seinen Nachbarn, die Zwischenzentralität (between-
ness centrality) als Anzahl der kürzesten Wege aller
möglichen Knotenpaare, die durch einen Knoten
führen, oder die Nähezentralität (closeness centrality), die die Länge dieser kürzesten Wege betrachtet.
Für Grad- und Zwischenzentralität sprechen hohe
Werte tendenziell für eine größere Bedeutung, wogegen für die Nähezentralität das Umgekehrte gilt,
vgl. [19].
Die folgenden Abschnitte betrachten zwei
Data-Mining-Ansätze genauer, die sich einerseits
auf die Inhalte (Sentiment Mining), andererseits auf die (Netzwerk-)Struktur von Social
Media (Community Mining) konzentrieren: Sentiment Mining wird genutzt, um Meinungen und
Stimmungen aus Social Media zu extrahieren.
Community Mining versucht, in der sozialen Netzstruktur eng zusammenhängende Gruppen zu
identifizieren.
Sentiment Mining
Sentiment Mining (Sentiment Analysis, Opinion
Analysis) [13] zielt darauf ab, mithilfe von NLP, linguistischen Methoden und Text Mining subjektive
Informationen aus textuellen Daten zu extrahieren. Es wird versucht, die Polarität (positiv/negativ)
eines Textes zu identifizieren. Die Polarität kann
auch genauer differenziert werden, z. B. ,,traurig“,
,,enttäuscht“ im negativen Fall, bzw. ,,fröhlich“,
,,glücklich“ im Positiven.
Die verwendeten Methoden basieren auf
Techniken des maschinellen Lernens, wie etwa
latente semantische Analyse und Support-VectorMaschinen, aber auch auf lexikonbasierten
Verfahren, die entsprechende Sentiment-Werte
für einzelne Worte enthalten und aggregiert den
Sentiment-Wert eines Textes bzw. einer Textpassage angeben. Zur Bewertung der Ergebnisse
werden die Vorhersagen mit einer menschlichen
Bewertung (meist mehrerer Personen) verglichen. Anwendungen reichen von der Einordnung
von Produktkommentaren bis zur Bewertung von
Blog-Einträgen.
Community Mining
Community Mining zielt auf die Auffindung und
Analyse (zusammenhängender) Subgruppen, Cluster oder Gemeinschaften ab, die in einem sozialen
Netzwerk ,,dicht“ miteinander verbunden sind.
Techniken zur Entdeckung von Communities
umfassen graph-basierte Ansätze [8], ClusteInformatik_Spektrum_35_2_2012
133
{ MINING SOCIAL MEDIA
ring aufgrund von Knoteneigenschaften oder
Pattern-Mining-Ansätze zur Optimierung einer
Community-Bewertungsfunktion [7]. Die Grundidee der Bewertungsfunktionen ist es, ein objektives
Bewertungskriterium zu finden, beispielsweise die
Anzahl von Verbindungen innerhalb der Community verglichen mit der statistisch ,,erwarteten“
Anzahl basierend auf allen verfügbaren Verbindungen im Netzwerk. Dabei werden die Communities
bevorzugt, die diese Bewertung lokal bzw. zusammen mit anderen Community-Einteilungen global
optimieren [7].
Die entdeckten Communities können anschließend für Empfehlungen, facettiertes Browsing,
oder zur Personalisierung von Systemen genutzt
werden, indem die Zuordnung zu einer Community explizit gemacht wird bzw. die Informationen
anderer Mitglieder der Community implizit zur Anpassung der Anwendung verwendet werden. Zur
Bewertung der Communities eignen sich manuell
erzeugte Testdaten und Nutzerstudien, die allerdings
in der Durchführung oftmals teuer sind. Bei datenbasierten Ansätzen, vgl. [17], werden verfügbare
Sekundärnetzwerke genutzt um einen relativen Vergleich von Communities durchzuführen; dies erlaubt
daher eine einfache und kosteneffiziente Bewertung
von Communities.
Anwendungen
Beispiele für Anwendungsfelder von Mining Social
Media sind intelligentes Monitoring, Reputationsmanagement, Profilbildung von Kunden,
Produktmanagement und Werbung, was in vielen
Social-Media-Diensten wie beispielsweise Twitter
oder Facebook umgesetzt wird. Hier kann etwa
Sentiment Mining zur Auffindung positiver oder
negativer Produktbewertungen genutzt werden,
um das Produkt- und Reputationsmanagement
durchzuführen. Community Mining kann zur Entdeckung impliziter Communities (z. B. ähnliche
Personen mit ähnlichen Interessen) genutzt werden.
Diese können dann zur Empfehlung interessanter
Kontakte oder zur Produktwerbung eingesetzt werden, um Produkte vorzuschlagen, die Personen mit
ähnlichen Interessen gekauft haben, oder solche,
die von Freunden gut bewertet wurden. Anbieter
können dazu eine Verknüpfung mit Social-MediaDiensten, z. B. Facebook, nutzen. Weiterhin bieten
sich Communities und soziale Netzwerkanalyse
zur Positionierung von Werbeanzeigen an, die auf
134
Informatik_Spektrum_35_2_2012
potenzielle Kunden zugeschnitten sind. Ebenso
können explizite und implizite Communities für
weitere vielfältige Empfehlungen verwendet werden,
je nach System und Ressourcentyp beispielsweise für
Produkte, Nachrichten, Bookmarks, Publikationen,
Artikel oder aktuelle Themen.
Eine konkrete Anwendung von Empfehlungen, eingebettet in intelligente Systeme, lässt sich
an den Conferator- und MyGroup-Systemen [5, 14]
illustrieren: Die Conferator-Software bietet Tagungsbesuchern die Möglichkeit, ihre sozialen
realweltlichen Kontakte zu verwalten und den
Konferenzbesuch effizienter zu organisieren. Zur
Verknüpfung der realen mit der Onlinewelt werden
aktive RFID-Tags eingesetzt, die u. a. Gesprächskontakte erfassen können. Das System ermöglicht z. B.
die Erstellung eines umfangreichen persönlichen
Profils, die Annotation eigener Kontakte mit Freitext
und die Zusammenstellung eines personalisierten
Tagungsprogramms. MyGroup arbeitet mit ähnlicher Technik im Umfeld von Arbeitsgruppen, z. B.
im Fachgebiet Wissensverarbeitung der Universität
Kassel. Conferator wurde bei verschiedenen Konferenzen z. B. bei der LWA 2010 in Kassel, bei der ACM
Hypertext 2011 in Eindhoven sowie der LWA 2011 in
Magdeburg erfolgreich eingesetzt.
Mithilfe der entstehenden Social-Media-Daten
werden in Conferator Empfehlungen zum Vorschlag interessanter Interaktionen mit Personen
gegeben, die ähnliche Interessen oder ein ähnliches ,,Vortragsbesuchsverhalten“ zeigen. Hierzu
werden der Kontaktgraph der Konferenzteilnehmer sowie verschiedene soziale Netzwerkund Publikationsdaten genutzt. Eine ExpertenEmpfehlungskomponente [14] realisiert eine
Vorschlagsfunktion für Softwareentwickler, basierend auf dem MyGroup Kontaktgraph bzw.
individuellen Gesprächen (Kontakten) sowie ihren
Checkins in Software-Versionskontrollsystem. Beide
Ansätze können zum schnelleren und gezielteren
Aufsuchen von Gesprächspartnern genutzt werden.
Die eingesetzten Techniken basieren auf den oben
beschriebenen Community-Mining-Methoden und
Zentralitätsanalyse-Techniken, die auch zur Analyse von (konferenzspezifischen) Rollen eingesetzt
werden können [6].
Die beschriebenen Anwendungen lassen sich
leicht auf andere Nutzungskontexte übertragen, in
denen die Entdeckung von Rollen und Communities
betrachtet wird, beispielsweise in kollaborativen
Systemen von Firmen, zur Analyse von SocialMedia-getriebenen Organisationsnetzwerken wie
Wikis oder in Versionskontrollsystemen [14].
Das Aufdecken von Prozessketten, die Identifikation von Ansprechpartnern und Experten und
die Optimierung von Workflows ermöglichen
zahlreiche interessante Anwendungsmöglichkeiten des Data Mining. Insgesamt spannt Mining
Social Media damit den Bogen zur Kollektiven
Intelligenz [15].
Ausblick
Mit der Vielfalt der Daten, die in Social-MediaSystemen gesammelt werden und der stetig weiter
wachsenden Nutzerzahl nehmen die Möglichkeiten
und Chancen der Analyse weiter zu. Neben erweiterter Personalisierung und Empfehlungen ergeben
sich viele weitere Anwendungen, die durch Data
Mining in Social Media ermöglicht werden. Mit
der Ausdehnung sozialer Medien auf mobile und
ubiquitäre Umgebungen, beispielsweise in Smartphones und Sensornetzwerken, kulminiert dies
im ,,Reality Mining“ [16], welches Data Mining auf
unterschiedlichsten ,,alltäglichen“ Sensordaten realisiert. Hier entsteht ein Spannungsfeld zwischen
technischen Möglichkeiten, Interessen der Nutzer
und ihrer Privatsphäre, insbesondere hinsichtlich
der Speicherung, Verknüpfung und Auswertung
persönlicher Daten zur Ausgestaltung und Bereitstellung personalisierter Dienste. Richtlinien
zur rechtlich korrekten Ausgestaltung von SocialMedia-Diensten [12], Anonymisierungsfunktionen
sowie datenschutzerhaltende (privacy-preserving)
Data-Mining-Methoden [9] stellen hier Ansätze
bereit, um entsprechende Datenschutzaspekte zu
berücksichtigen.
Mit hinreichenden Maßnahmen zum Datenschutz ermöglicht Mining Social Media damit
letztendlich die Entdeckung neuen Wissens und ein
besseres Verständnis von Kommunikation, Interaktionen und kollektiven Prozessen, um Umfeld und
Handlungen der Anwender zu begleiten und sie im
Alltag bei der Umsetzung ihrer Ziele zu unterstützen.
Literatur
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
http://twitpic.com/135xa#, letzter Zugriff 31.12.2011
http://twitter.com/#!/UlrichKelber/status/1892741071, letzter Zugriff 31.12.2011
http://goo.gl/S7rNf, letzter Zugriff 31.12.2011
http://www.facebook.com/press/info.php?statistics, letzter Zugriff 23.1.2012
Atzmueller M, Benz D, Doerfel S, Hotho A, Jäschke R, Macek BE, Mitzlaff F,
Scholz C, Stumme G (2011) Enhancing social interactions at conferences. it – Inf
Technol 53(3):101–107
Atzmueller M, Doerfel S, Mitzlaff F, Hotho A, Stumme G (2011) Face-to-face contacts during a conference: communities, roles, and key players. In: Proc
MUSE 2011, Athens, Greece
Atzmueller M, Mitzlaff F (2011) Efficient descriptive community mining. In: Proc
24th Intl FLAIRS Conference, AAAI Press
Girvan M, Newman MEJ (2002) Community structure in social and biological networks. Proc Natl Acad Sci USA 99:7821–7826
Großkreutz H, Lemmen B, Rüping S (2010) Privacy-preserving data mining.
Informatik-Spektrum 33(4):380–383
Heidemann J (2010) Online social networks. Informatik-Spektrum 33(3):262–271
Kaplan A, Haenlein M (2010) Users of the world, unite! The challenges and opportunities of social media. Bus Horizons 53(1):59–68
Krause B, Lerch H, Hotho A, Roßnagel A, Stumme G (2010) Datenschutz im Web
2.0 am Beispiel des sozialen Tagging-Systems BibSonomy. Informatik-Spektrum
35(1):1–12
Liu B (2010) Sentiment analysis and subjectivity. Handbook of Natural Language
Processing. Chapman & Hall
Macek BE, Atzmueller M, Stumme G (2011) Profile mining in CVS-logs and faceto-face contacts for recommending software developers. In: Proc IEEE SocialCom
Malone TW, Laubacher R, Dellarocas C (2009) Harnessing crowds: mapping the
genome of collective intelligence. MIT Center for Collective Intelligence
Mitchell TM (2009) Mining our reality. Science 326:1644–1645
Mitzlaff F, Atzmueller M, Benz D, Hotho A, Stumme G (2011) Community assessment using evidence networks. In: Analysis of Social Media and Ubiquitous Data,
Springer
Stehlé J, Voirin N, Barrat A, Cattuto C, Colizza V, Isella L, Régis C, Pinton J-F, Khanafer N, Van den Broeck W, Vanhems P (2011) Simulation of an SEIR infectious
disease model on the dynamic contact network of conference attendees. BMC
Med 9(87)
Wassermann S, Faust K (1994) Social Network Analysis. Cambridge University
Press
Informatik_Spektrum_35_2_2012
135