|Title:||Engineered IT management automation along task analysis, loops, function allocation, machine capability catalog|
|Committee:||Bernhard Neumair, Alexander Knapp, Christian Böhm|
PhD project description
Limited predictability of emergent behavior, lack of transparency and service guarantees, neglected security requirements, and a general lack of control are properties which are currently associated with self-managing systems, leading to scepsis and limited user acceptance of such systems in productive environments of service providers. At the same time, management automation systems are expected to have desirable features such as model-based automation, autonomic operations according to policies, robustness, scalability, and resilience to errors by human operators, all of which promise lower operational costs. Marketing efforts by hardware, software and solution vendors stimulate business and therefore scientific interest in current and future "self-managing" products and services. Current systems are not completely managed automatically, instead they are managed systems equipped with certain management automation routines that deal with certain operational aspects. Depending on both the importance of these aspects for the functionality of the whole system and the maturity of the used management automation mechanisms, the systems can be associated with some level of automation, between complete dependence on and obedience to external human control and complete automation. To find out, to what extent both, properties of managed systems and management automation mechanisms, can be combined, I investigate "management automation systems", that show the conflict of external management and management automation in such systems.
Criteria are introduced which - in combination with metrics and general questions on design issues - will be compiled into an evaluation sheet to assess current systems with self-management capabilities. The idea is to have a look at existing systems from information and communication technology, as well as systems from the aeronautical, space, automotive, medical, and industrial sector and to assess whether design principles or patterns can be applied to the management of IT components and services. I know, that the research domains mentioned before have a certain tradition in systems engineering as well as existing systems that replace or assist human operators/users. There, *adjustable autonomy* (switching between various autonomy level offered by a resource or service) is one promising approach to the issues raised above with an adjustable trade-off between human influence and supervision on the one hand and efficient automated decision making on the other.
Once sufficient data has been collected, the results of these evaluations shall serve as a basis for comparing systems with self-management capabilities, to come up with ideas to test and benchmark such systems and to create a classification scheme that groups existing solutions according to common properties. The main focus benchmark is the resulting "level of autonomy" and its relationships to a) the management actions, that still need to be carried out by humans and b) the resulting system performance. I believe, that the ratio of management input and system performance output will be a major decision factor when purchasing or subscribing to future IT resources and services.