Dr. Mikhail Panshenskov

St. Petersburg State University, Russia

Date:
Time:
Place:

Thursday, January 31^st 2013
11:00
Library, Argentinierstrasse 8 / E184-1, 3rd floor

This talk has been cancelled

Abstract:

In the presence of limited resources constant computing power growth requires effective utilization of computing resources. The problem of effective scheduling becomes highly relevant. Numerous heuristics (like min-min, xsufferage, genetic algorithms etc.) provide relatively good solution for a NP-hard problem of scheduling. Although most of the published heuristics have two issues. First of all, many of such algorithms do not consider inter-processor communication which becomes highly important in modern computing. Secondly, many scheduling heuristics do not guarantee the quality of scheduling whatsoever. Based on models of C.Leiserson and V.Vl. Voevodin the author builds a new mathematical model of scheduling in distributed systems with active data transfers. The author considers different types of independent tasks and different types of distributed systems and for all the scenarios presents worst case estimates of quality of scheduling as a function of task and distributed environment parameters. These estimates demonstrate the influence of communication and sizes of independent tasks on the quality of scheduling. For instance, the Author shows how the connection latency plays important role in the scheduling. Also the author has deployed a distributed system on small cluster of 20 nodes and demonstrated applications of the method to estimate the quality of scheduling. The methods can be applied to calculate the quality of scheduling for different present cases including 1)scheduling of independent tasks in MapReduce and 2)scheduling in the presence of horizontal scaling (like in elastic cloud EC2).

Paul Grefen & Remco Dijkman

Eindhoven University of Technology

Date:
Time:
Place:

Thursday, January 17^th 2013
11:00
Library, Argentinierstrasse 8 / E184-1, 3rd floor

Abstract:

In this presentation, we address the topic of support for real-time control in complex, multi-modal logistics chains. We elaborate control models and show how these can be mapped to information system infrastructures. The presentation consists of two parts. In the first part, Paul Grefen presents the development of an information-based control model for supply chains. He shows how this control model can be mapped to shared IT infrastructure, such as service clouds. In the second part of the presentation, Remco Dijkman presents the GET Service European project, which aims at a practical implementation of a specific version of the control model around an integrated service bus. In this project, aggregation of large volumes of diverse real-time, transport-related information is used to provide input for both static and dynamic planning of complex logistics processes. The ultimate goal is to support on-the-fly, drag-and-drop reconfiguration of logistics processes using the parallel availability of multiple transport modalities (i.e., using synchro-modality).

at the 10^th International Conference on Service-Oriented Computing, Shanghai 2012

Former DSG member Martin Pinzger accepted the Full Professor position of Software Engineering at the University of Klagenfurt, Austria

Prof. Schahram Dustdar receives the IBM Faculty Award 2012.
TU Wien News Page (german)

Artikel in der "Presse" vom 22.08.2012

Seite 24, Seite 25, Seite 26

Die TU Wien forscht für Städte der Zukunft und entwickelt Ideen und Lösungsansätze.
Smart City Forschung @ TUU Wien.

Artikel im Kurier vom 13.6.2012

Doctoral College "Adaptive Distributed Systems"

more Information

Alessio Gambi

Univ. of Lugano, Switzerland, [Homepage]

Date:
Time:
Place:

Thursday, January 12^th 2012
14:00
Library, Argentinierstrasse 8 / E184-1, 3rd floor

Abstract:

In the context of Cloud computing, Service Providers (SP) can access resources "on-demand" and dynamically scale their elastic applications; in this way, SPs can reduce the need of system over-sizing and provide guarantees about applications QoS. To properly achieve this, SPs must decide on when, what and how to scale their systems.

Dynamic changes in operating conditions, and the complexity of the relations between system behaviors and system configurations may reduce the effectiveness of statically defined solutions. Model based self-adaptive controllers seem to be more promising solutions as they were originally proposed to deal with complex systems, emerging behaviors, and dynamic environments, and because they can autonomously learn from past experience and adapt to the actual operating conditions.

In this talk, I discuss the design of self-adaptive controllers based on a particular black-box surrogate model, named Kriging model. Kriging models are the mathematical tools that controllers use to capture the relations between the QoS of the virtualized systems, the incoming workload and the system configuration, and to support their planning activities; the models are maintained up-to-date by the controllers during the runtime activities.

Info Brochure
from Pacific Controls

 

Prof. Frank Leymann

Universität Stuttgart, Germany

Date:
Time:
Place:

Wednesday, Nov 30^th 2011
15:00
Library, Argentinierstrasse 8 / E184-1, 3rd floor

Pacific Controls and the Vienna University of Technology are developing an "internet of things". Automatic communication between electronic devices and infrastructures can save money and energy.

TU Vienna press release:
german english

 

Christian Haas

Karlsruhe Institute of Technology (KIT), Germany

Date:
Time:
Place:

Wednesday, October 5^th 2011
16:00
seminar room Argentinierstrasse

Abstract:

In order to run a platform like a Social Cloud, a certain quantity of resources is necessary to facilitate the platform itself and the services that enable the basic operation of the platform. For example, because users can share and trade resources using different (economic) protocols, the matching of demand to supply results in the calculation of one or more allocations, and requires a form of computational infrastructure to facilitate this.

Hence, this talk discusses how such infrastructure resources can be provided by the users of the Social Cloud itself, rather than from a central entity. An economic co-operative sharing model of users is formulated, where the platform itself and its functionality are provided and owned by the users. In addition, the efficiency and performance of such a model depending on incentive schemes and users' characteristics is studied.

Rui Oliveira

University do Minho, Braga, Portugal

Date:
Time:
Place:

Thursday, April 14^th 2011
14:15
EI1, Gusshausstrasse 25

Abstract:

This guest lecture is part of the master level lecture "Advanced Distributed Systems".

It may be of particular interest for PhD students working on distribution aspects, as consensus is *the* fundamental problem of distributed systems and many proofs exists under what conditions consensus can be deterministically solved - and when not (e.g., two army/coordinated attack problem, byzantine generals, muddy children/jealous amazons, etc...).

The lecture will explain the basic problem, sketch the FLP proof idea and motivate the idea of failure detectors.

Dr. Stefan Schulte

TU Darmstadt

Date:
Time:
Place:

Friday, April 1^st 2011
13:00
Library, Argentinierstrasse 8 / E184-1, 3rd floor

Abstract:

After a brief description of my research background, this talk will focus on my former work on process reliability in Service-oriented Architectures. The creation of business processes by composing loosely coupled services to (complex) workflows is one of the major application scenarios of Service-oriented Computing. As a result of the ongoing success of Web service technologies, services can be purchased from external providers. In a future "Internet of Services", functionally equivalent services which provide different cost and quality levels will be available and can be selected by a service consumer in order to execute the different steps of a workflow. Furthermore, the existence of functionally equivalent services allows replacing a service if it does not meet its guaranteed service levels. In this talk, I will present work originating from the BMBF-funded project SoKNOS (Service-Oriented ArchiteCtures Supporting Networks of Public Security) which addresses the selection of services based on user-defined constraints, the monitoring of the workflow execution, and the re-planning of service-based workflows (if necessary).

Press Release at CeBIT 2011

Jährlich wird der MiA-Award an Frauen mit Migrationshintergrund für hervorragende Leistungen und Erfolge in und für Österreich vergeben. Die diesjährige MiA für Wissenschaft & Forschung erhielt Ivona Brandic für ihre bemerkenswerten wissenschaftlichen Leistungen.

Die Universitätsassistentin stammt ursprünglich aus Bosnien-Herzegowina und kam aufgrund des Kriegsausbruches mit 15 Jahren nach Österreich.
Ivona Brandic studierte Wirtschaftsinformatik an der TU Wien sowie Universität Wien und begann ihre wissenschaftliche Karriere am Institut für Scientific Computing der Universität Wien. Seit 2007 ist sie am Arbeitsbereich Distributed Systems des Instituts für Informationssysteme tätig.

Für die MiA wird aus jeweils drei nominierten Kandidatinnen eine Preisträgerin in den folgenden Kategorien ermittelt: Wissenschaft & Forschung, Wirtschaft, Humanitäres & Gesellschaftliches Engagement, Kunst & Kultur, Sport.

Professor Dr. Jörn ALTMANN

Technology Management, Economics and Policy at the College of Engineering Seoul National University

Date:
Time:
Place:

Monday, 21 February 2011
16:00
Library, Argentinierstrasse 8 / E184-1, 3rd floor

Prospects for a Collaboration on Evaluating Adaptive SOAs

Dr. Jan Sudeikat

HAW Hamburg / VSIS group at the University of Hamburg

Date:
Time:
Place:

Tuesday, 15th February 2011
10:00
Library, Argentinierstrasse 8 / E184-1, 3rd floor

Dr. Kamal Bhattacharya

Senior Manager at IBM Research India

Date:
Time:
Place:

Monday, 31 January 2011
11:00
Library, Argentinierstrasse 8 / E184-1, 3rd floor