IMPROVING INFORMATION SYSTEM'S ROBUSTNESS BY FAST RECONFIGURATION

System reconfiguration is often the only means of preserving the continuity of business-critical services after a host crashes or a successful penetration attack is perpetrated. As a consequence, a fragment of the system infrastructure (hosting the services) has to be isolated to prevent escalation of damages caused by the fault. The other system hosts need to take over the functions normally performed by the affected servers. The paper proposes a complete approach to manage this reconfiguration: identifying the service components that need to be relocated, the hosts that they should be deployed to, predicting the impact that it has on system performance. It describes an efficient approach, based on network simulation, for assessing the response times of services after reconfiguration, taking into consideration the resource consumption interactions between the co-located services.