Thorsten Schöler

An observer/controller architecture for adaptive reconfigurable stacks

In this paper, we discuss the necessity of new observation and control structures for organic computing systems starting from the basic contradiction between bottom-up behaviour and top-down design. An Observer/Controller architecture serves the purpose to keep emergent behaviour within predefined limits. As an illustration, a framework for reconfigurable protocol stacks is introduced, which contains an agent-based monitoring framework as well as a reconfiguration manager. After describing a TCP/IP protocol stack implementation, based on the framework, similarities between the introduced framework and the Observer/Controller architectural pattern will be pointed out.

Design, implementation and validation of a generic and reconfigurable protocol stack framework for mobile terminals

We introduce a modular and reconfigurable software framework for protocol stacks implemented in platform independent manner. Simulation tools useful for software validations are introduced and a new distributed, three-staged procedure for validation of protocol stack software is proposed. Assertion-based virtual prototyping (based on nonresident assertions), utilising simulation of hardware software co-systems as well as software probes containing code-resident assertions are used in the proposed validation process.

Automatic flow classification using machine learning

Network standards are moving toward the quality-of-service (QoS) networking. Differentiated services (DiffServ) QoS model is adopted by many recent and upcoming networks standard. Applications running on these networks can specify suitable service classes to their connections or flows. The flows are then treated according to their service classes. However, current Internet applications are still designed based on best-effort scheme and, therefore, cannot benefit from QoS support from the network. An automatic flow classification framework, which can automatically classify non QoS-aware flows or legacy flows, has been proposed in our earlier work [2]. In this paper, we extend our framework by introducing new features that can be effectively used to classify legacy flows. The simplicity of these features allows the data to be collected in real-time. No packet-level data are required. Furthermore, the framework is evaluated using multiple data sets from different users. The results show that our framework works extremely well in general and it can be operated independently from any applications, networks or even machine learning algorithms. Average correctness up to 98.82% is achieved when the framework is used to learn and classify unseen flows from the same user. Cross-user classifications yield average correctness up to 74.15%.

First steps towards organic computing systems: monitoring an adaptive protocol stack with a fuzzy classifier system

Protocol stacks for small devices like mobile phones have to fulfill multiple functions in different environments. This has lead to highly complex solutions with alternative stacks. It is more economical to construct new stacks from building blocks on the fly depending on the current requirements. Such an adaptive architecture is an example of an organic computing system. In this paper we show how a proposed monitoring architecture for adaptive protocol stacks fits into the observer/controller pattern of organic computing. The protocol stack framework and the observer/controller pattern are briefly described. Subsequently the observer/controller pattern is mapped onto the framework architecture. Two implementations for monitoring agents are described, a finite state machine solution and a fuzzy classifier approach. The fuzzy classifier system is described in detail. It is shown that a finite state machine and a standard Markovian classifier system can be generalised into a non-Markovian learning classifier system, combining the advantages of both parents