Additionally, the following publications are partially used in this thesis, although they are not directly related to the main contributions: 1. Modeling Elasticity Trade-Offs in Adaptive Mixed Systems, Muhammad Z.C. Candra, Hong-Linh Truong and Schahram Dustdar, The 22nd IEEE International WETICE Conference (WETICE 2013), Adaptive Computing (and Agents) for Enhanced Collaboration (ACEC) track, June 17-20 , 2013, Hammamet, Tunisia [45]. 2. Virtualizing Software and Human for Elastic Hybrid Services, Muhammad Z.C. Candra, Rostyslav Zabolotnyi, Hong-Linh Truong, and Schahram Dustdar, Advance Web Services, (c)Springer-Verlag, 2012 [46].
1.5
Scopes of Work
Computing systems comprising humans and machines as compute units cover a wide spectrum with different characteristics. In Section 2.1, we discuss the landscape of the state of the art hybrid human-machine computing. Our work presented in this thesis focuses on HCSs with the following characteristics. An HCS may consist of human-based compute units, software-based compute units, and/or thing-based compute units. Our work is motivated by issues raised due to the inclusion of human-based compute units in a computing system, such as discussed in Section 1.3. Therefore, this thesis focuses on systems where human-based compute units are involved. Systems focusing only on software-based compute units, and/or thing-based compute units, e.g., [47], although applicable, are less relevant in this thesis. Our work assumes the existence of a coordinator role in the system, which manages tasks creation, distribution and execution in an automated manner. Depending on the problem domain, such a coordinator can be manifested in different forms, such as a process engine, an application middleware, etc. This excludes, for example, systems with ad-hoc interactions, e.g., for online collaborative content creation [13], and crowdsourcing systems where tasks are manually posted, and results are rather manually evaluated, e.g., [48]. Furthermore, our work necessitates the feasibility for applying a formal task formulation on the system, with respect to the multidimensional aspects of tasks, such as roles and activities that compute units could take, quality constraints, and the dependencies between task activities. Our main contributions presented in this thesis require such information for controlling quality-aware provisioning of compute units, monitoring constrained metrics, and analyzing task dependency structures for reliability analysis. Such a task formulation is discussed in Section 3.3. 8