Lehrstuhl für Technische Informatik - Rechnernetze

Investigation of Using Automated Planning Methods in IT Service Recovery

Organizational:

This thesis is a part of cooperation between MNM Team (LMU) and HP Research Lab Bristol, UK. Student has chance to travel to HP Research Lab Bristol and work with top class scientists and engineers there during completion of her/his thesis.

Short Description(Kurze Beschreibung):

The availability of IT services plays an essential role for today's IT service providers. In order to improve the service availability, it is necessary for a service operator to reduce the service downtime and recover the impacted service efficiently in case of an outage. This goal can be achieved either by increasing the reliability of service components or reducing the time for service recovery with the well-established recovery plans.

During the service outages, service operators have to make series of decisions based on the available information (e.g. necessary actions, available resources for recovery, service level specification and past recovery history etc.) to recover the impacted service to its normal operational state. As the IT services and their underlying infrastructures are getting more complicated, it is a challenging task for operators to make timely decisions on recovery procedures correctly. In order to alleviate the above mentioned problem, it is helpful to develop a mechanism to assist the operator to generate recovery plans of actions automatically according to the available information.

Automated planning is a well-established and widely used technology to support decision-making process. It has been successfully implemented in the areas like manufacturing, space exploration (NASA's Mars Exploration Rover) and project management, etc. Algorithmically, a planning problem has as input a set of possible courses of actions, a predictive model for the underlying dynamics, and a set of policies to guide plan searching. A planning problem thus involves deciding what actions are necessary to bring the current state to desired state. The output or solution is one or more courses of actions that satisfy pre-determined requirements. To generate plans for service recovery, different aspects must be taken into consideration regarding the complexity of today's IT environment, for example, planning under uncertainty, planning with incomplete domain information, etc.

Tasks(Aufgaben):

• Investigate the applicability of different automated planning methods in IT service recovery scenarios;
• Evaluations of applicable methods according to metrics like complexity, correctness and efficiency;
• Determine essential components to support the planning processes.

Requirements(Anforderungen) :

• Java, Python or comparable OOP. Knowledge of logical programming (Prolog, LISP) is ideal but not required;
• Knowledge of artificial intelligence, especially automated planning, is advantageous;
• Knowledge of ITIL would be appreciated.

Topic Assignment(Aufgabensteller): Prof. Dr. H.-G. Hegering

Duration(Dauer der Diplomarbeit): 6 Months

Number of Candidates(Anzahl Bearbeiter): 1 or 2

Advisor(Betreuer):

• Feng Liu
• David Trastour (HP Lab Bristol, UK)