Autonomic Management of Large-Scale Critical Infrastructures

Critical Infrastructures include facilities, services and installations essential for the functioning of a society and economy. Such infrastructures are generally of a very large scale spanning cities, whole countries, or even crossing international borders. Examples of such Large-Scale Critical Infrastructures (LSCI) are electricity, water and gas supply, transportation, and health service [8]. There are many aspects to the management of LSCI, including security, fault tolerance, availability, and reconfiguration. The particular aspect of LSCI management with which we are concerned and which we view as a grand challenges in autonomic computing is optimization of their performance in changing conditions. Not all circumstances in which these systems will operate can be predicted so it is not possible to fully define their behaviour at design time. Even for the known operating conditions, with hundreds or thousands of nodes, it is infeasible, if not impossible, to define correct behaviour for all combinations of conditions on all nodes. Critical infrastructures need to adapt to various changes in load, both sudden ones and repeated load patterns. They need to optimize their performance with respect to multiple, often conflicting or highly dependent, policies with different levels of priority (high, low), affecting different parts of the systems (local, regional, global), either continuously, or in certain cirumstances (sporadically). In our work, Urban Traffic Control (UTC) is used as an exemplar of a LSCI. A UTC systems consist of hundreds of dependent nodes (traffic lights) that need to coordinate their behaviour to deal with optimizing general traffic throughput, prioritizing emergency vehicles and public transport, as well as adapting to any surges of traffic in particular areas in case of public events or accidents.