Research

Research

Statement

Socio-technical systems experience more than ever before the complexity of a new multidisciplinary information era. By definition, societal, ecological and technological development is distributed and dynamic: People, locations, time events, resources and our natural environment are all inter-related, inter-dependent and continuously changing. Capturing and managing this information using centralized computing systems in not a scalable solution. The design, control and management of these systems can be advanced by decentralized computing systems. The roadmap for the intelligence, robustness and sustainability of large-scale complex networks such as the Internet and the Smart Power Grid should move towards their integration with self-managed computing systems.

By definition, societal, ecological and technological development is distributed and dynamic: People, locations, time events, resources and our natural environment are all inter-related, inter-dependent and continuously changing.

The study, design and prototyping of decentralized self-managed systems require one or more virtual meta-levels of information that abstract the infrastructures and its physical assets: Overlay networks. Such decentralized environments are built by autonomous software agents that compete and cooperate to achieve their goals by interacting in a peer-to-peer fashion. Furthermore, these local interactions should be coordinated by the agents themselves to meet global application objectives. As central coordination is not always an option, agents should be self-organized, self-optimized and self-configured within overlay network topologies built and maintained by dynamic methodologies and mechanisms. Tree, clustered and super-peer topologies supported by gossip-based epidemic protocols and bio-inspired algorithms make fundamental operations such as information dissemination, information lookup, aggregation and decision-making possible and cost-effective in decentralized environments. As these methodologies are often application-dependent and integrated in the application-level, designing these mechanisms as decentralized middleware services can provide application capabilities that are more generic, modular and customizable.

As central coordination is not always an option, agents should be self-organized, self-optimized and self-configured within overlay network topologies built and maintained by dynamic methodologies and mechanisms.

Self-management does not exclude system actors and domains from system decisions and control. Instead, self-management is a multidisciplinary and collective process that builds the required synergies and adaptations between different system actors from different domains with opposing objectives. The Smart Power Grid, as a socio-technical system, is a representative application example that brings together energy providers, energy consumers, operators, policy makers and a complex physical infastructure. Collectively and beyond their different objectives, these stakeholders should cooperate and compete within a self-managed system that should support the robustness of the Smart Power Grid and the energy sustainability of society.

…self-management is a multidisciplinary and collective process that builds the required synergies and adaptations between different system actors from different domains with opposing objectives.

For more information, explore my research projects and publications.

A high-leve research overview: specialization and applications.

Specialization

1. Peer-to-peer and Agent-based Overlay Networks

Dynamic gossip-based overlay networks

Clustered and tree overlay networks

Super-peer overlay networks (scale-free networks)

2. Self-* Properties

Self-organization of overlay networks and distributed information

Self-optimization of resource utilization in favor of distributed applications

Self-configuration of local agent behavior to meet global system objectives

3. Network and System Complexity

Graph properties of distributed networks

Adaptation of autonomous agents

Emergence from remote network interactions

4. Design and Engineering Distributed Systems

Middlewares

Services

Prototyping

Applications

1. Self-management of Networks

Aggregation

Load-balancing

Information dissemination

2. Smart Power Grid

Stabilization of energy consumption

Shifting of energy consumption

Composition of embedded control systems

3. Distributed Multimedia

Multicasting

File sharing

Networking games