[Back]

J. Cik:
"Software-defined Elastic Systems Management in Multi-cloud Environments";
Supervisor: H. Truong; Institut für Informationssysteme, Distributed Systems Group, 2015; final examination: 2015-06-02.

Elastic systems are characterized by the ability to expand and contract in the cloud environment, through dynamic de/allocation of resources, based on their requirements. These requirements can be described, with respect to the internal structure of the elastic
system, as a trade-off between resource, quality and cost. In order to make such systems feasible, a platform specialized on development and operation of elastic systems is needed. Within this platform, supporting services for deployment, monitoring, testing and elasticity control handle all the tasks for the additional aspect of elasticity, leaving the developer´s hands free to deal with the system´s domain-specific logic. Moreover, extending such a platform over multiple existing clouds results in an integrated multicloud
environment.
The challenge is to manage elastic systems in this platform throughout their entire lifecycle. This includes coordination of elastic platform services, while providing traceability of actions executed by the services as well as by human users. Furthermore, it is
necessary to support extensibility of the platform with new services and offer a uniform interface for interaction with the platform.
This thesis proposes a framework for management of elastic systems in the above described platform. The resulting management system is based on loose coupling of elastic platform services through an observable event-driven communication. The exchange of events is centered around a lifecycle of an elastic system, which serves for
coordination of the involved components. The lifecycle reflects platform´s use cases and characteristic relations between elastic platform services. The thesis further proposes components of the management framework, including management of dynamic elastic
platform services, as well as, recording and analytics of events. Last but not least, a prototype is implemented for evaluation of the framework and of the introduced concepts.