Author: Andreas Dittrich Proceedings of the Joint Workshop of the German Research Training Groups in Computer Science, Algorithmic synthesis of reactive and discrete-continuous systems (AlgoSyn), Dagstuhl, Germany, May 31 – June 2, 2010 Download: extended abstract

Disasters striking in inhabited areas pose a significant risk to the development and growth of modern societies. The impact of any disaster would be severe. In case a disaster strikes, fast and safe mitigation of damages is important. Information and communication technology (ICT) plays a crucial role in helping reconnaissance and first response teams on disaster sites.

Most rescue teams bring their own network equipment to use several IT services. Many of these services (e.g., infrastructure, location, communication) could be shared among teams but most of the time they are not. Coordination of teams is partly done by pen and paper-based methods. A single network for all participating teams with the possibility to reliably publish, discover and use services would be of great benefit.

Despite the participating teams and course of action being different on every site, described service networks display certain common properties: They arise spontaneously,
the number of nodes and their capabilities are subject to high fluctuation, the number and types of services are also fluctuating strongly and there is no global administrative configuration.

Because of these properties all network layers involved would need to be configured automatically. Based on the Internet Protocol (IP) — the only well-established global networking standard — a number of mechanisms promise to automatically configure service networks. In disaster management scenarios, where various services are critical for operation, mission control could benefit from these mechanisms by getting a live view of all active services and their states. It needs to be investigated if and how they are applicable.

Given an ad-hoc, auto-configuring service network, how and to what extent can we guarantee dependability properties such as availability, the ability to perform in the presence of faults (performability) and ultimately the ability to sustain certain levels of availability or performability (survivability) for critical services at run-time?

The goal of this dissertation is to provide a comprehensive dependability evaluation for such heterogenous and dynamic service networks. A run-time dependability cycle is being embedded into the network. In this cycle, the network is constantly monitored.
A distributed service discovery layer provides network-wide service presence monitoring. This will be extended to provide monitoring for availability and performability assessment. Based on monitoring data, dependability properties are evaluated at run-time. The survivability of critical services can be estimated by calculating the expected availability or performability with a given fault model. If necessary, adaptation measures are triggered which in turn can cause the monitoring to be reconfigured. Even if no adaptation is possible, run-time awareness of critical states is already a huge benefit. This cycle is the base of a self-aware adaptive service network.

Categories: Publication, Research and Education
Tags: Adaptivity, Availability, Client-server systems, Dependability, Dissertation, Distributed computing, IP, Network, Service dependability, Service network management, Survivability