[Back]

P. Zeppezauer:
"Virtualizing Communication for Hybrid and Diversity-Aware Collective Adaptive Systems";
Supervisor: H. Truong; Institut für Informationssysteme, Distributed Systems Group, 2014.

Hybrid and Diversity-Aware Collective Adaptive Systems (HDA-CAS) form a broad class of highly distributed systems comprising a number of heterogeneous human-based and machinebased computing (service) units. These units are required to perform tasks on their own and in cooperation with other units to solve complex problems. Therefore, they typically interact and collaborate in an ad-hoc manner. These units form dynamic adaptive collectives, which are subject to constant change. Whenever possible, the collectives are allowed to self-orchestrate, using familiar collaboration tools and environments.
The flexibility of these collectives makes them suitable for processing elaborate tasks. At the same time, building a system to support diverse types of communication types in such collectives
is challenging, because the way how human-based and machine-based units communicate differs fundamentally. To be able to use both in a hybrid system, the actual way of communication between the units has to be virtualized and handled in the system independently of the actual type of the communication participants allowing for a uniform communication between applications of the HDA-CAS platform and individual service units.
In this thesis, the fundamental communication challenges for HDA-CAS are addressed and requirements, and properties of communication in a HDA-CAS are formulated. This thesis discusses these problems and presents a concept of how to virtualize the communication with
service units and collectives. Therefore, the notion of service units is extended and the concept of communication adapters is discussed. Furthermore, this thesis presents a design of a middleware which uses the concepts for virtualizing the communication between, within and
among collectives and service units. The middleware is able to handle numerous, intermittently available, human and machine-based service units, and manage the notion of collectivity transparently to the programmer. A prototype implementation for validation and evaluation purposes is also provided.